1881.] 



the Thermoelectric Quality of Metals. 



401 



shown graphically in a number of diagrams. To apply a load, remove 

 parfc of it, re-apply it, and continue the application, forms a wide hut 

 closed, or nearly closed, loop on the " on " curve. Similarly, to remove 

 load, stop during the removal, and replace part, and then go on again 

 with the removal, forms a loop of a like kind on the " off " curve. 



The curves got during the gradual removal of load always show a 

 negative maximum, even though the applied load has been less than 

 that required to pass the negative maximum during loading. 



The effect of beginning to reload a stretched wire which has been 

 unloaded is at first feebly positive, passes a positive maximum, becomes 

 negative, passes a negative maximum, and finally, if the wire does not 

 break too soon, becomes once more positive before rupture takes place. 



The leading characteristics of the cyclic action mentioned above are 

 sufficiently obvious from the diagrams, but can scarcely be described 

 verbally, except at great length. The curves show a lagging of thermo- 

 electric effect when the stress is changed. It is proved that this lag- 

 ging of effect is not a time action, for it is shown to be independent of 

 the rate of increment and decrement of the stress. Moreover, the 

 thermoelectric value associated with any load (not exceeding the 

 elastic limit), arrived at in any manner, is constant so long as the load 

 is constant. 



One very remarkable feature of the curves is that the first effect of 

 reversal from loading to unloading, or from unloading to loading, is to 

 continue the same hind of thermoelectric change which has been going 

 on just before the reversal takes 'place. For example, if the wire had 

 been becoming negative when loading was stopped and unloading 

 began, it continued to change towards negative during the very first 

 part of the subsequent unloading. This law appears to be general. 



As a consequence of the cyclic action it follows that there may exist 

 between a stress and its associated thermoelectric effect any relation 

 which can be expressed by a point within the wide area enclosed 

 between the most distant " on " and " off " curves. To attempt, there- 

 fore, to assign a relation irrespective of the preceding states and 

 changes of stress would be altogether futile. 



Instances are given of "molecular reminiscences " of previous stress- 

 actions, exhibiting themselves by modifying the form of the curves in 

 the next succeeding experiments with the same wire. 



It is shown that mechanical vibration greatly reduces, if it does not 

 wholly destroy, the distinction between the "on" and "off" curves of 

 thermoelectric quality and stress, if the wire be kept in a state of vibra- 

 tion during the application and removal of the load. 



It is also shown that two widely different values of thermoelectric 

 quality, got by reaching the same stress in two different ways (namely 

 by addition of load from zero and by partial removal of load from a 

 high value), become almost equal to each other when the wire is 



