June 2, 1892] 



NATURE 



i03 



After a couple of hours' standing the leakage through the 

 india-rubber and at the joints could be measured. The 

 amount of the leakage found in the first two hours was 

 usually negligible, considered as an addition to a globe- 

 ful of hydrogen, and the leakage that would occur in the 

 hours following would (in the absence of accidents) be 

 still smaller. If the test were satisfactory, the filling 

 would proceed as follows : — 



The electric current through the generator being 

 established, and the furnace being heated, any o.xygen 

 that might have percolated into the drying tubes had 

 first to be washed out. In order to do this more 

 effectively, a moderate vacuum (of pressure equal to about 

 I inch of mercury) was maintained in the tubes and up to 

 the regulator by the action of the pump. In this way the 

 current of gas is made very rapid, and the half-hour 

 allowed must have been more than sufficient for the pur- 

 pose. The generator was then temporarily cut off, and a 

 high vacuum produced in the globe connection and in the 

 blow-off tube, which, being out of the main current of gas, 

 might be supposed to harbour impurities. After this the 

 pump would be cut off, the connection with the generator 

 re-established, and, finally, the tap of the globe cautiously 

 opened. 



The operation of filling usually occupied from two to 

 three hours. When the gas began to blow off under an 

 excess of pressure represented by about half an inch of 

 mercury, the blow-off cistern was lowered so as to leave 

 the extremity of the tube free. For two minutes the 

 current of gas from the generator was allowed to flow 

 through, after which the generator was cut off, and the 

 globe left in simple communication with the atmosphere, 

 until it was supposed that equilibrium of pressure had been 

 sufficiently established. Doubts have at various times 

 been felt as to the interval required for this purpose. If too 

 little time is allowed,there will remain an excess of pressure 

 in the globe, and the calculated weight of the filling will 

 come out too high. On the other hand, an undue pro- 

 longation of the time might lead to a diffusion of air back 

 into the globe. In a special experiment no abnormal 

 weight was detected after half an hour's communication, 

 so that the danger on this side appeared to be small. 

 When the passages through the taps were free from 

 grease, one or two minutes sufficed for the establishment 

 of equilibrium, but there was always a possibility of a 

 partial obstruction. In the results to be presently given, 

 four minutes were allowed after the separation from the 

 generator. It may be remarked that a part of any 

 minute error that may arise from this source will be 

 eliminated in the comparison with oxygen, which was 

 collected under like conditions. 



The reading of the barometers and thermometers at 

 the moment when the tap of the globe was turned off 

 took place as described in the former paper. The 

 arrangements for the weighings were also the same. 



In the evacuations the process was always continued 

 until, as tested by the gauge of the Toppler after at least 

 a quarter of an hour's standing, the residue could be 

 neglected. Here, again, any minute error would be 

 eliminated in the comparison of the two gases. 



In the case of oxygen, the errors due to contamination 

 (even with hydrogen) are very much diminished, and 

 similar errors of weighing tell very much less upon the 

 proportional agreement of the final numbers. A com- 

 parison of the actual results with the two kinds of gas 

 does not, however, show so great an advantage on the 

 side of the oxygen as might have been expected. The 

 inference appears to be that the individual results are 

 somewhat largely affected by temperature errors. Two 

 thermometers were, indeed, used (on opposite sides) 

 within the wooden box by which the globe is surrounded, 

 and they could easily be read to within -^' C. But in 

 other respects, the circumstances were untavourable, in 

 consequence of the presence in the same room of the fur- 



NO. I 1 79, VOL. 46] 



nace necessary to heat the copper. An error of ± o*^*! C. 

 in the temperature leads to a discrepancy of i part in 

 1 500 in the final numbers. Some further elaboration of 

 the screening arrangements actually employed would 

 have been an improvement, but inasmuch as the circum- 

 stances were precisely the same for the two gases, no 

 systematic error can here arise. The thermometers were, 

 of course, the same in the two cases. 



The experiments are grouped in five sets, two for 

 oxygen and three for hydrogen. In each set the work 

 was usually continued until the tap of the globe required 

 re-greasing, or until, owing to a breakage or to some 

 other accident, operations had to be suspended. 



The means are as follow : — 



Hydrogen. 



1891. 



Weight. 



July 



September 

 October .. 



Mean 



gram. 

 0-15808 

 0-15797 

 0-15804 



temp., F. 



60 



Globe 

 temp., C. 



Corrected 



gram. 

 0-158056 



0-157950 

 0-158040 



16 0-158015 



The means here exhibited give the weights of the two 

 gases as they would be found with the globe at 12° C, 

 and the barometers at 60^ F. and at 30 inches. The 

 close agreement of the mean temperatures for the two 

 gases shows how little room there is for systematic error 

 dependent upon imperfections in the barometers and 

 thermometers. But the results still require modification 

 before they can be compared with the view of deducing 

 the relative densities of the gases. 



In the first place, there is a systematic, though minute, 

 difference in the pressures hitherto considered as corre- 

 sponding. The terminal of the blow-off tube is 33 inches 

 below the centre of the globe at the time of filling. In 

 the one case this is occupied by hydrogen, and in the 

 other by oxygen. If we treat the latter as the standard, 

 we must regard the hydrogen fillings as taking place 

 under an excess of pressure equal to \% of the weight of 

 a column of oxygen 33 inches high ; and this must be 

 compared with 30 inches of mercury. Hence, if we 

 take the sp. gr. of oxygen under atmospheric conditions 

 at 00014, and that of mercury at 13-6, the excess of 

 pressure under which the hydrogen was collected is as a 

 fraction of the whole pressure 



0-0014 , 



— ~^— = o'oooioo ; 

 13-6 ' 



and 0000106 ,X 0-158 = 0000017. This, then, is what we 

 must subtract from the weight of the hydrogen on account 

 of the difference of pressures due to the gas in the blow- 

 off tube. Thus 



H= 0-157998, = 2-51724. 

 But there is still another and a more important cor- 

 rection to be introduced. In my former paper it was 

 shown that when the weighings are conducted in air the 

 true weight of the gas contained in the globe is not given 



