and Reflexion of Gas Molecules. 367 



highly polished. It was cemented in position with sealing- 

 wax, with the reflecting surface at the centre of the bulb. 



The bulb was wrapped in cotton, which was kept soaked 

 with liquid air throughout the experiment, which lasted about 

 thirty minutes. 



A photograph of the deposit is shown on Plate VII. fig. 2; 

 and it is evident that the clear zone is absent, the deposit 

 coming quite down to the plane of the reflecting surface. 



Measurements of the thickness of the film for different 

 angular distances from the line normal to the centre of 

 the reflecting surface, made by observing photometrically 

 the transmission of red light by the film, showed that the 

 cosine law was obeyed to within the probable error of the 

 experiment. 



During the progress of the investigation, however, the 

 exceeding ingenious method devised by Knudsen was com- 

 municated by him to me in a letter, and as this method was 

 so obviously superior to that above described, the quanti- 

 tative investigation of the law of reflexion was abandoned in 

 its early stages. 



Knudsen's method, which was published in the Annalen 

 der Physik in the early part of 1916, is based upon the fact 

 that for reflexion from a spherical surface under certain con- 

 ditions a film of uniform thickness will be formed if the 

 reflexion from each element of the spherical surface follows 

 the cosine law. This he found to be the case. 



The general subject of condensation and reflexion as a 

 function of the temperature was next investigated. 



It is apparent, from the experiment with the bulb which 

 I have described, that cadmium vapour, at least when highly 

 attenuated, will not condense as a homogeneous film on a 

 glass wall at room temperature, whereas it will on a surface 

 of metallic cadmium. 



In the vicinity of the heated bead of the metal we may 

 have a metallic film produced, but here the nature of the 

 deposit is very variable, depending on the condition of 

 the glass surface. Black splotches sometimes form, granular 

 deposits of white metal and mirror surfaces may alternate, 

 and in fact no two tubes act alike. 



Iq the bulb, however, the conditions are different, and the 

 phenomena can be reproduced at will. 



I first attempted to get some idea of how many collisions 

 with the wall were possible without condensation, by sub- 

 stituting for the bulb a glass tube bent at a number of right 

 angles, as shown on Plate VII. fig. 3. The end of the tube 



