478 Scientific Intelligence. 



M due alone to the warm mercury vapor when dynamical equili- 

 brium obtains between the saturated vapor and the gas. Curves 

 were next drawn with temperatures of the vapor as abscissae and 

 the corresponding values of k as ordinates. On the same diagram 

 curves were plotted which show the dependence of k on t when 

 the condition p x + P 1 = jo, 4- P 2 was assumed to hold. This 

 equation is the one generally used in theoretical discussions of 

 the process of diffusion, and it states that the sum of the partial 

 pressure p of the gas and P of the mercury vapor is constant. 

 The last set of curves fall closer to the axis of k than the set first 

 mentioned, thus showing that the conditions in a vacuum differ 

 appreciably from those which prevail at higher pressures, such as 

 0-004 mm to 0-04 n)m . 



The next step in the argument consisted in the calculation of 

 the coefficient of diffusion D from Stephan's law and from data 

 obtained from experiments in vacuo. The values of D came out 

 so nearly constant as to show complete agreement between theory 

 and experiment. Hence, the lack of coincidence of the two sets 

 of curves is not due to theoretical error but corresponds to reality. 

 Consequently, the relation p 1 + P i = p 2 4- P 2 needs correction for 

 finite pressures. On hydrodynamieal grounds Gaede next works 

 out the correction term e in the equation p } + P 2 = p 2 -f P 2 + e. 

 This term is too complicated to be given here. However, it 

 involves the coefficient of diffusion P, the radius of the diffusion 

 tube, the coefficient of slip, the coefficient of viscosity, and the 

 several partial pressures. It follows from the form of the func- 

 tion e that this term can only become negligible under certain 

 special conditions. On the other hand, it is quite possible for e 

 to have appreciable values. That such is the case is tantamount 

 to answering the question suggested at the beginning of this ab- 

 stract. In brief, it has now been shown both practically and theo- 

 retically that it is possible to attain, by means of a mercury pump, 

 pressures lower than the vapor pressure of the mercury involved. 

 Also, these pressures may be correctly indicated by a MacLeod 

 gauge. The fallacy in the argument that total pressures lower 

 than that of saturated mercury vapor at room temperature 

 (0 # 0013 mm ) cannot be obtained with a mercury pump arises from 

 the tacit assumption that e = at finite pressures. 



Some idea of the principle of action of the final form of diffu- 

 sion pump, which the preceding considerations enabled Gaede to 

 design and perfect, may be obtained from the following brief 

 description of an illustrative pump of very simple construction. 

 A cylinder of unglazed porcelain, with the upper " base " of the 

 same material, is surrounded at the top and sides by a larger 

 non-porous cylinder which in turn is joined at its top, by means 

 of a suitable tube, to a steam generator. A metal cylinder, closed 

 at the top, projects up inside the porcelain vessel and closes the 

 lower end of the latter completely. Cold water is caused to cir- 

 culate in the metallic cylinder. The space between the coaxial 

 cylinders communicates with a bulb and a barometer tube. The 



