210 MKSSRS. A. SCHUSTER AND G. HEMSALECH ON 



questions connected with tins mutter, as we are only concerned with the bearing of 

 our experiments on the main subject of our research. 



The curved lines of the oscillatory discharges may serve as a basis for the calcu- 

 lation of the molecular velocities, and if this is done, higher values are obtained 

 than those we have derived from our experiments with the complete apparatus. 

 Aluminium and magnesium gave again, however, the highest velocities near the pole, 

 but that of the zinc molecules exceeded considerably the velocities found for cadmium. 



There is here, however, a considerable difficulty in the interpretation of the result, 

 as without the prism it is impossible to separate the different lines, and, what is 

 perhaps more important, we do not get the effect of the first discharge at all, as that 

 is hidden behind the dense luminous column of the straight air discharge. 



If a photograph be taken of the spark on the rotating film, the image is always 

 drawn out most near the middle of the spark, and we obtain images such as fig. 31. 

 It is not quite easy to see why the metallic particles should remain luminous in the 

 centre of the spark longer than near the pole, unless there is some inflow of cold air 

 from the poles inwards. Such an inflow might be produced through the effects of 

 rarefaction produced by the initial heating of the air by the spark. 



8. Sources of Error. 



Different photographs, obtained exactly in the same way, sometimes differ con- 

 siderably in the value they give for the molecular velocity, and there is little doubt 

 that the chief cause of error lies in the fact that the discharge is not a straight line 

 parallel to the slit, but takes place in irregular curves, and, as appears already on 

 Dr. FEDDEBSEN'S photographs, the successive oscillations of the same discharge take 

 sometimes very different courses. When we first began to work we used smaller 

 capacities, and our results were more irregular, because the course of the discharge 

 was more erratic. The large capacity acts in the direction of making the path of the 

 discharge straighter, unless the sparking gap becomes too great. If the molecular 

 stream is a straight line, our calculations are based on the further assumption that 

 the image of the point of divergence falls exactly on the slit. If that point is at 

 a distance h, measured perpendicularly to the slit, the molecules would have to 

 traverse a distance \/y* + A 2 , if the projection of the distance along the slit is y. 

 Calling V the velocity of the molecular stream, v that of the photographic film, it 

 follows from x = ~Vt and vt = \/y- + A 2 that 



V 2 ^ - t>y = AV. 



x and y are proportional to the coordinates of the spectroscopic lines as measured on 

 the film, and the curve represented by the equation is a hyperbola. 



The error introduced is such that a constant velocity would appear as one infinitely 

 large close to the pole and gradually diminishing. If k is small, the hyperbola would 

 soon coincide with its asymptote, and in that case the error in the molecular velocities 



