Pooi.E — On ihe Convection of Heat in Vertical Wider Columns. 209 



flowing through a flat metal worm immersed in it. In most of the experiments, 

 the water, before entering the worm, passed through a coil of metal tubing 

 immersed in the lower part of the large water-bath E. 



A and B were enclosed in an inverted bell-jar D, the intervening space 

 being filled with granulated cork, and B further screened by a Dewar vessel, 

 as shown in the figure. Metal ballast in the bottom and a few small weights 

 on top of the cork kept B floating vertically in the water-bath, which was 

 enclosed in a barrel F, the intervening space being filled with granulated 

 cork. The woim tube in C was independently supported. 



The difference of temperature between the ends of A, ranging from a small 

 fraction of a degree to about 10 degrees, was read by a thermo-couple, 

 whose junctions were enclosed in thin glass tubes passing through the annular 

 corks into B and C respectively, so that the junctions were level with and 

 close to the ends of A. The temperatures of C and E were read by a pair 

 of accurate mercury thermometers, one being independently supported with its 

 bulb close to the upper thermo-junction, and the other floating in E with 

 its bulb rather below the level of B. No stirring was used, as it was desirable 

 to work with the water in A as quiescent as possible, but even so the water 11. 

 E was generally at a very nearly uniform and constant temperature throughout 

 a test. 



This arrangement was almost entirely made up of apparatus which chanced 

 to be available. A thermostat would obviously have been an improvement, 

 but a rather complicated system of temperature control would have been 

 necessary, as it was desirable to work below 20° C, and considerable quantities 

 of heat had to be got rid of. Even if C and E were maintained at constant 

 temperatures, the mean temperature of the water in A would vary with the 

 heat supplied to its base, so, to keep the latter temperature constant, C should 

 be cooled as the power supply was increased. In view of the essentially 

 rough character of the tests such a complication did not appear to be justified. 



The thermo-couple consisted of 5 pairs of iron and constantan wires, the 

 total resistance being 85 ohms, and was connected through a Morse telegraph 

 key to a galvanometer of resistance about SCO ohms, the key being arranged 

 so that the galvanometer was normally short-circuited. The couple was inserted 

 in the circuit by depressing the key, which was enclosed in a wooden box and 

 operated by a long vulcanite handle to reduce thermal eifects. The zero was 

 re-set, when necessaiy, by moving the galvanometer lamp, the scale being 

 fixed. In this way thermal effects at the galvanometer were almost completely 

 eliminated. It was found that leakage from the battery supplying the heating 

 current might cause a shift of. a few divisions in the zero, but did not 

 appreciably -affect the deflection produced by depressing the key. This was 

 tested by reversing the heating current, and also by momentarily disconnecting 

 the battery. In the later experiments the zero shift was entirely eliminated 

 by earthing one end of the heating winding, which was electi-ically connected 

 to the vessel B. Identical results were obtained whether this earth connexion 

 was made or not. 



The couple and galvanometer were calibrated throughout the scale of 500 

 divisions by comparison with the two mercury thermometers, these, in turn, 

 being compared with a standard. The deflection per degree varied from 32-9 

 scale divisions for small deflections to 320 for large ones, and was sensibly 

 independent of the mean temperature of the couple over the range (10° C. to 

 25° C.) covered. 



The current and the P.D. across the heating coil were both measured, and 

 it was found that over a large part of the range of power used, the resistance 



