Glass, and Metal Wires subjected to Constant Pull. 519 







a-S 



Pull in grains. 



log*. 



logt' 



14725 



•18 



25-0 



1467*5 



•35 



27-2 



1462-5 



•51 



28-5 



1457*5 



■66 



29-5 



1452-j 



•88 



360 



1448-5 



100 



35-3 



1444-5 



117 



36-1 



14405 



1-29 



35-4 



1436-5 



1-43 



353 



1430-5 



1-68 



36-2 



1427-5 



1-77 



35-8 



1412-5 



2-32 



360 



1384-5 



3-16 



34-3 



Subsequent experiments have shown that after about a day 

 the decay of stress becomes much more rapid until S becomes 

 very small, and then S gradually approaches a zero limit. 

 In the experiment of which the observations are given the 

 pull after twelve months had fallen to 22 grams. 



Stretching by different proportions of the unstretched 

 length shows that b is proportional to the stretch. 



All the ordinary assumptions which one can make, such as 

 considering the stretch as consisting of elastic and viscous 

 portions, seem to lead to the conclusion that the stretch 

 should be a function of e~ t and not of log t. 



Consequently it was thought that the amplitude of the up 

 and down oscillations of a weight on the end of a length of 

 indiarubber bandage might obey some other law than the 

 usual one. Observations showed, however, that the usual 

 law is obeyed, i. e. that the amplitude decreases in geometrical 

 progression as the time increases in arithmetical progression. 



In order to give some idea of the effect of alternations of 

 temperature in the laboratory in producing errors, the tem- 

 perature coefficients of the iudiarubber were found while the 

 indiarubber was in tension. The coefficient of expansion on 

 cooling proved to be much larger than the coefficient of con- 

 traction on heating. Consequently the alternations of tem- 

 perature in the laboratory would produce a slow creep on 

 their own account: thus when the creep is very slow the 

 effect of the alternations of temperature maybe much greater 

 than the true logarithmic creep. In all cases it was found 

 that when the creep became very slow, e.g. a few days after 

 the load was put on, it was a great deal faster than the 

 logarithmic law would warrant. 



