August 23, 1900] 



NATURE 



407 



case. The position of the sphere was read in the usual way by 

 scale and telescope. The time of swing of this little sphere was 

 120 seconds. 



A larger quartz sphere, 6 6 cm. diameter and weighing 

 400 gnis. , was fixed at the lower end of an axis which could be 



Jfotor, 



ToAccuTivt 



^?Ti' 



Governor] 

 Lead, Flu-irheeL 





::,f, -^^f% , , v 



Fig. 9 — Experiment on directive action of one quart2 crystal on another. 



turned at any desired rate by a regulated motor. The centres 

 of the spheres were on the same level and 59 cm. apart. On 

 the top of the axis was a wheel with 20 equidistant marks on its 

 rim, one passing a fixed point every 1 1 -5 seconds. 



It might be expected that the couple, if it existed, would have 



Fig. 10. — Upper curve a regular vibration. Lower curve a disturbance 

 dying away. 



the greatest effect if its period exactly coincided with the 120 

 second period of the hanging sphere — i.e. if the larger sphere 

 revolved in 240 seconds. But in the conditions of the experi- 

 ment the vibrations of the small sphere were very much damped, 

 and the forced oscillations did not mount up as they would in a 



\^ 



freer swing. The disturbances, which were mostly of an im- 

 pulsive kind, continually set the hanging sphere into large 

 vibration, and these might easily be taken as due to the re- 

 volving sphere. In fact, looking for the couple with exactly 

 coincident periods would be something like trying to find if a 

 fork set the air in a resonating jar vibrating 

 when a brass band was playing all round it. 

 It was necessary to make the couple period, 

 then, a little different from the natural 120 

 second period, and, accordingly, we revolved 

 the large sphere once in 230 seconds, when 

 the supposed quadrantal couple would have & 

 period of 1 1 5 seconds. 



Figs. 10 and il may help to show how this 

 enabled us to eliminate the disturbances. Let 

 the ordinatesof the curves in Fig. 10 represent 

 vibrations set out to a horizontal time scale. 

 The upper curve is a regular vibration of range 

 + 3, the lower a disturbance beginning with 

 range ± 10. The first has period i, the second 

 period i 25. Now cutting the curves into lengths 

 equal to the period of the shorter time of vibration, and arranging 

 the lengths one under the other as in Fig. II, it will be seen 

 that the maxima and the minima of the regular vibration always 

 fall at the same points, so that, taking 7 periods and adding up 

 the ordinates, we get 7 times the range, viz. ±21. But in the 

 disturbance the maxima and minima fall at different points, and 

 even with 7 periods only, the range is from + 16 to - 13, or less 

 than the range due to the 

 addition of the much smaller 

 regular vibration. 



In our experiment, the 

 couple, if it existed, would 

 very soon establish its vibra- 

 tion, which would always be 

 there and would go through 

 all its values in 115 seconds. 

 An observer, watching the 

 wheel at the top of the re- 

 volving axis, gave the time 

 signals every 11 -5 seconds, 

 regulating the speed, if neces- 

 sary, and an observer at the 

 telescope gave the scale read- 

 ing at every signal, that is, 

 10 times during the period. 

 The values were arranged in 

 10 columns, each horizontal 

 line giving the readings of a 

 period. The experiment was 

 carried on for about 2\ hours 

 at a time, covering, say, 80 

 periods. On adding up the 

 columns, the maxima and 

 minima of the couple effect 

 would always fall in the same 

 two columns, and so the addi- 

 tion would give 80 times the swing, while the maxima and minima 

 of the natural swings due to disturbances would fall in different 

 columns, and so, in the long run, neutralise each other. The re- 

 sults of different days' work might, of course, be added together. 

 There always was a small outstanding effect such as would be 

 produced bya quadrantal couple, but its effect was not always in the 

 same columns, and the net result of about 350 period observations 

 was that there was no 115 second vibrationwjf more than I second 

 of arc, while the disturbances were sometimes 50 times as great. 



The semicircular couple required the turning sphere to revolve 

 in 115 seconds. Here, want of symmetry in the apparatus 

 would come in with the same effect as the couple sought, and 

 the outstanding result was, accordingly, a little larger. 



But in neither case could the experiments be taken as show- 

 ing a real couple. They only showed that, if it existed, it 

 was incapable of producing an effect greater than that observed. 

 Perhaps the best way to put the result of our work is this : 

 Imagine the small sphere set with its axis at 45° to that of the 

 other. Then the couple is not greater than one which would take 

 Si hours to turn it through that 45° to the parallel position, and 

 it would oscillate about that position in not less than 12 hours. 



The semicircular couple is not greater than one which would 

 turn from crossed to parallel position in 4^ hours, and it would 

 oscillate about that position is not less than 17 hours. 



1^ 



/=^ 



Fig. II. — Results of 'superposition of 

 lengths of curves in Fig. 10 equal 

 to the period of the regular one. 



NO. 1608, VOL. 62] 



