L ' e l 4-9 8-6 14-2 18-4 22-2 25-5 32-3 35*9 39'0 45-6 488 56-35. 



464 Mr. A. Griffiths on Diffusive Convection. 



and H being closed. The taps E and H were then opened, 

 and a rush would doubtless occur along CD. After some 

 time, perhaps 5 or 10 minutes, the taps E and H were closed, 

 and the ink added. 



The chief data are as follows : — 



Diameter of A or B =1*09 centim. approximately. 



Diameter of capillary =0*019 centim. approximately. 



Height of top of B above top of A =0'08 centim. 



Height of bottom of B above bottom of A = 0*292 centim. 



Length of A = 4*622 centim. 



Length of B = 4*410 centim. 



The motion of the centre of the liquid in the capillary is 



indicated in the table below. 



Hours from start ... 1 18^43 66 90 



Motion of centre 

 in centim 



If a curve be drawn with hours as abscissas, and the dis- 

 tance moved as ordinates, a regular diminution of the flow 

 is indicated, the velocity tending to a steady value. The 

 average velocity in the last three days was 2*5 centim. per day. 



Calculations made on the principles indicated in the previous 

 section give about *8 centim. per day as the velocity. 



The observed velocity is thus three times the calculated. 

 As the apparatus when originally put together was only 

 intended to give qualitative results, it would be premature 

 to discuss the cause of the difference *. 



After the above observations had been made, the tube B 

 was lowered with respect to A ; the flow along the capillary 

 from left to right was stopped, and there was probably a flow 

 in the opposite direction. (The ink is not adapted for 

 for detecting slight backward motions.) 



A second apparatus made by Mr. Rixon was fitted up, and 

 a solution of common salt employed. Currents were produced 

 in the direction indicated by theory. This new apparatus is 

 capable of delicate adjustment, and experiments with it are 

 still in progress. To the author the interest lies not so much 

 in the confirmation of the theory as in the possibility of a 

 physical method of determining the coefficient of diffusion. 



* Note added July 15th, 1898. Later experiments indicate that the 

 difference between the calculated and observed velocities in the above 

 was due to the continual increase of density of the solution owing to 

 evaporation of water at the surface. On covering the surface with oil, to 

 prevent evaporation, the observed and calculated values are found to be 

 close enough to prove the substantial truth of the theory. Thus in one 

 experiment, the observed motion of the index, at the end of a fortnight, was 

 2-3 centim. per day, whereas theory indicates that the final velocity should 

 be 2-1 centim. per day. 



