High Vacuum Pumps.--A revolution in high vacuum tech nique was brought about in 1905 when Gaede introduced his rotary mercury pump. In principle the pump, which is illus trated in fig. 6, is something like a gas meter run backwards. It consists of a porcelain drum W, rotating on an axis A, mounted in an iron vessel GG, the front of which is closed by a glass plate B, as shown in the upper figure. The drum is divided into two parts by a vertical wall T, the portion on the right of this wall enclosing a space V which communicates through the tube R with the vessel to be evacuated. The portion to the left of the wall is divided into three compartments by walls of the shape shown in the lower figure. The pump is filled with mercury up to the level qq. When the drum rotates in the direction of the arrow the compartment which communicates with V by the hole f, is filled with gas from the receiver, while the gas in the compartment drawn from the receiver just before, is dis placed by the mercury and ultimately passes into the space be tween the revolving drum and the casing. The action is contin uous. It is, however, clearly necessary that the pressure prevail ing in the space between drum and casing shall not exceed a few millimetres or so of mercury : the pump must therefore be run in connection with a fore-pump of the type described, which first reduces the pressure to a figure at which the rotary mercury pump can be put into operation, and subsequently removes, as long as necessary, the gases displaced by the mercury pump. The fore pump is connected to the tube 52, the other branch being con nected to a rough mercury gauge.
With mercury pumps of this type a pressure as low as mm. of mercury, as measured with the McLeod gauge, can be attained. There is, however, always the pressure of the mercury vapour itself, about .00i mm. of mercury at ordinary room temperature, unless special precautions are taken to trap the vapour between pump and receiver. Further, the pump cannot deal with condens ible vapours, such as water vapour, for the decrease of volume of the chamber during the expulsion of the gas leads to condensation of the saturated vapour in the drum. The pump needs a large mass of mercury. It has, however, the great advantage that it can be stopped without deterioration of the vacuum already attained, which cannot be done with other types of pump to be described. The speed is about ioo cc. per sec. over a wide range, but begins to fall off rapidly after a pressure of io microbars or so has been reached, being one quarter this value at •i microbar. This pump is still widely used in the laboratory.
In 1912 Gaede invented a pump working on an entirely different principle, the so-called molecular pump. This takes advantage of certain properties of low pressure gases demonstrated by Knud sen. If we consider a passage of rectangular cross-section, one wall of which (or two opposite walls) is moving in its own plane in the direction of the axis of the passage, then molecules striking the moving wall acquire a forward component of velocity equal to that of the wall. If the gas is at a pressure such that the mean free path is very small compared to the transverse dimension of the passage, this ordered velocity, imposed on top of the random velocities of the molecules, is, as we pass away from the surface, gradually dissipated into a chaotic velocity distribution by impact among the molecules. The motion is that considered in the ordi
nary treatment of viscous fluids, and a difference of pressure can be maintained between the two ends of the passage which is proportional to the speed at which the wall is moving. As the pressure is diminished the viscosity remains independent of the pressure until this becomes so low that the mean free path is comparable with, or large compared to, the transverse dimensions of the passage. (See KINETIC THEORY.) At very low pressures the experimental effects can be accounted for on the assumption that the molecules striking a stationary wall are thrown back in directions independent of the angle of incidence, the number of molecules streaming from the wall in a given direction obeying the same simple cosine law as governs the intensity of the light emitted from a glowing plate. As far as the direction of the "reflected" molecules is concerned, the gas behaves as if it were condensing on the wall and evaporating again. If the wall which the molecules strike is moving, the velocity of the wall must be added to that of the molecules (at any rate if the pressure is below .00i mm. of mercury; at higher pressures the effect at the wall is somewhat complicated by the formation of a skin of gas, which has the effect of producing an apparent slip at the walls. (See Gaede, Annalen der Physik, 41, 289, 1913.) There will be effectively a convection of the molecules by the moving surface. Theoretical considerations on lines worked out by Knud sen and Gaede shows that with the low pressures contemplated the ratio of the pressures at the two ends of the passage is fixed by the speed of the wall, instead of the difference of the pressures, as at higher pressures.
Suppose now a deep and narrow groove cut in the periphery of a rotating cylinder, which fits closely into a casing, bored with two close holes communicating with groove. Between the two holes a tongue projects from the casing into the groove, with which it makes a close fit. If the cylinder rotates quickly, the gas in the groove is between close moving walls, and molecules will be carried away from the side of the tongue from which the walls are receding, and carried to the side of the tongue which they are approaching. A vacuum will tend to be produced on the former side, so that of the two borings mentioned one will be the low pressure side and the other the high pressure side of the pump. The two sides are in permanent communication by the groove, but molecules can only pass from one to the other by traversing practically the whole circumferential distance against the direction of the moving walls. With practicable speeds of rotation the ratio of the pressures at the two ends of the groove turns out to be sufficiently large to permit a very efficient high vacuum pump to be constructed on the principle of adding the velocity of the walls to the molecular velocity at low pressure.