DowLiNG — Steady and Turbulent Motion in Gases. 389 



The critical velocity for a gas in a particular tube is proportional to rj/p. 

 Thus the critical velocity would be seven-fold greater for hydrogen than for 

 air. It would be difBcult to measure so great a velocity. Thus our choice 

 remains between the otlier two. Oxygen diifei's little from air. Carbon 

 dioxide was accordingly used in the experiments. 



In the experiments with carbon dioxide two gasometers were employed. 

 The gas flowed from one to the other through the experimental tube GABC. 

 As it was not convenient to take elaborate precautions against leakage, the 

 following device was adopted : — A second wire (X) was attached to the 

 drum V of the first gasometer, and carried over tlie pulleys as shown in 

 fig. 1. This wire passed in an exactly similar fashion over the pulleys 

 of the second gasometer, and was attached to the drum thereof. Thus if 

 no balance weights were hung from the original wire R, the gas in both 

 the gasometers would be under identical pressures. Wlien the weight was 

 hung on its wire, it diminished the " effective " weight of V, and the other 

 drum pressed on the gas in the second gasometer. Thus at one end of tlie tube 

 there was a sliglit excess, and at the other end a slight defect of pressure. 

 At no point would there be any serious difference from atmospheric pressure. 



In performing an experiment, the gas was drawn, of course, in the direction 

 ABC; and when the gasometer at G was full, the gas was driven back 

 down the tube ready for anotlier series of numbers. The speeds were 

 adjusted as before by either ^altering the driving weiglit W, or tlirottliiig 

 the flow by the tap T. Preliminary experiments with this double gasometer 

 arrangement showed that it worked even better than the original method. 

 Steadier and more concordant readings were obtainable at all times. This 

 was probably due to the fact that draughts and other external conditions 

 did not influence the blast of air inside tlie tube. In consequence of this.,, 

 some of the previous curves for air which liad not been regarded as 

 satisfactory, were repeated with this apparatus. These are indicated in 

 Table I by tlie remark " tico gasometers.^' 



Being satisfied by tlie experiments with air of the working of the new 

 arrangement, one of tlie gasometers was filled with carbon dioxide; and 

 after " washing out " the tube-system and the other gasometer with tlie 

 gas, the filled gasometer was connected to the tube system ABC. The 

 method of filling precluded tlie possibility of much admixture of air, and 

 a small quantity would produce oulj' a small effect on the viscosity, whicli 

 for a mixture is given by the formula 



""- V,,+ V, ^^■> 



due to Graham. (Wliere »)'" is the coefficient of viscosity of a mixture of 



