AND RATIO OK THEIR ATOMIC WKKillTS. 75 



was closed, and only wlu'ii, in closing, the crack was turned towards the globes. 

 This determination ought fairly to l)e rejected, althougii there was no means of 

 knowing, at the time, that leakage was actually taking place. 



There were some accidents which might make the weight of a litre of hydro- 

 gen appear too high. In bi'eakiug otf tiie [)oint ?i\ Fig. 'J'k an unnoticed fragment 

 might V)e lost. The asbestos in h, if not [)roperly adjusted, might permit the escape 

 of a globule of the melted metal. If the mt)istiire absorbed by the asbestos during 

 the construction of the tube was not removed during the first exhaustion of the 

 space f/ h, Fig. 25, it would be given off with the hydrogen. Either of these 

 would make the weight of the hydrogen expelled from the tube appear too great. 

 It was thought that the first experiment of series iv. was in error from the first of 

 these causes, though the evi<lence was not so conclusive as to lead to the instant 

 abandonment of the experiment. 



Five experiments were made before tlie proper conduct of the experiment was 

 mastered. The mishaps were various, and perhaps do not need to be detailed. 

 The observations are given, though they are not used in obtaining the final mean. 



26. lIYDR0(iKN KY NEW JIETIIOD. SECOND Al'l'AKATUS. 



The determinations of series iii. were made as soon as the apparatus was put 

 together and the preliminary experiments finished, and while the apiiaratus was in 

 perfect order. The stopcocks showed no leakage even in a somewhat protracted 

 test. But after an absence during the summer, much tiouljle \vas experienced, of 

 which the true cause was not easily learned. Leakage appeared, sometimes slow, 

 but sometimes rapid or capricious, so that many experiments were abandoned at 

 different stages. Fresh lubrication of the stopcocks not helping nuich, they were 

 surrounded with rubber capsules holding glycerine. This stopped the leakage 

 lone euoush to eet the results of series iv., but the leakage then became so capricious 

 that sometimes an action of the pump \vhich ought to have given a vacuum of one 

 millionth left one oi' two thousandths instead. The apparatus was broken up and 

 replaced with one which should be proof against leakage ; some time afterwards 

 the nature of the difficulty was learned. The crack in the key of the stopcock 

 caused the small leak. The rubbei- capsules holding glycei-ine, by continued pres- 

 sure, grjidually loosened the key of one stopcock, admitting glycerine, then watei-, 

 and finally air; hence the larger and capi'icious leakage. 



27. HYDROGEN BY NEW METHOD. SECOND AIM'ARATHS. CAPACITY OE ULOBES. 



It will have been noticed that the stopcocks of the three globes, Fig. 23, 

 were useful only during the measurement of the capacity of the connecting tubes. 



