ON A QUARTZ TIIKKAD (UUYITY BALANCE. 255 



nf IH'.M. We now had a balance much worse than the one that had been 

 broken : the thread appearing to show a viscous yielding in the wrong direction. We 

 attributed this to the shellac, so in January, 1894, an experimental thread was 

 mounted on yet another balance. This thread was drawn down so as to be thick in 

 the middle and at both ends, with the view of reducing the stress intensities in the 

 shellac. The results of this experiment were such as to lead us to think the abnormal 

 behaviour of the portable Iwilance was to be traced to the thread or lever, and not to 

 the shellac. However, by February the balance seemed to have settled down, and it 

 was again taken to Armidale. On this occasion, however, the results were quite 

 disappointing, and the cross wire images were so difficult to observe that it was clear 

 that a new mirror or method of observing was necessary. The viscous subsidence 

 from whatever cause arising was also unsatisfactory, so that we decided to mount a 

 fresh thread, choosing now a much finer one. For two months of incessant work we 

 struggled with fine threads, finally mounting what we thought a very good one in 

 April, and this was then observed till July, when we have the following note : " This 

 thread, the very finest we have had with a very small lever, has been most unreliable. 

 It seems from its behaviour (especially when the lever becomes unstable) that the 

 centre of gravity of the lever has been moving relatively to the thread, perhaps the 

 fine thread has been moving through the shellac."* A defect in the arrester rendered 

 this possible. 



In consequence of this observation a separate experiment on another balance was 

 made from May till September, using a soldered thread. We were led by this experi- 

 ment to solder up a fine thread to the portable balance and to mount on it a soft copper 

 lever. On August 21st, after continued observation, we were forced to the following 

 conclusion : " From the general appearance of the observations it is clear that the 

 readings are becoming less with lapse of time or the thread requires more twist to 

 bring the lever to its sighted position the effect to be anticipated . . . The readings 

 are, however, irregular . . . The thread is not to be compared with others we have had 

 whose variations were within a degree, whereas this thread has altered its reading by 

 ten degrees. The thread is so fine that the centre of gravity of the lever must be 

 extremely close to it, so that the very smallest change in the lever or attached mirror 

 will greatly affect the reading. The very fine threads have not been a success." 

 As a result of this experience we abandoned the lever and mirror in favour of a 

 microscope, and also brought the arrester to its present form. During September 

 and October, 1894, several threads were mounted, and these all broke before we 

 could get a series of observations. These persistent failures led us to examine the 

 threads more closely than we had done hitherto, both as to uniformity and freedom 

 from air bubbles. Threads of the following diameters all broke: '00126, '0015, and 



* We are now inclined to think that a good deal of the difficulty which we attributed to the imperfect 

 elasticity of the threads was in reality due to bad means of attachment, imperfect spring, and bad 

 temperature estimations. 



