GRAVITATION AS AFFECTED BY TEMPERATURE. 
387 
teats may be considered crucial. Curiously, it was found that though these were the 
purest silver obtainable from Messrs. Johnson, Matthey and Co., they were both in 
slight degree permanently magnetised. This is in itself an interesting fact. The 
torsion system used here is, in this case, a most sensitive magnetometer and when a 
N. pole of a bar magnet was brought outside the vacuum, slightly above one silver 
sphere, it repelled it, the corresponding scale reading being 4 mm. ; whereas when 
the N. pole was below, there was a like attraction. This permanent magnetism was 
removed with some difficulty by the passage of a current through a coil round the 
vacuum tube, until at last the scale reading was reduced to the small amount of 
0'2 mm. for each sphere. Thus the permanent magnetism was reduced to l/20th of 
its former amount. Several days later the spheres, m, m, were again tested and were 
found to remain demagnetised. Probably the permanent magnetism was originally 
produced when the heating coils round the vacuum tube were excited during 
production of the vacuum. (See § VI., paragraph 9.) 
The above demagnetisation of m, m occurred on date 20th August, 1915, after 
which, as is seen from Table IV., several full experiments were performed yielding 
results similar to those found before demagnetisation. From this one may deduce 
that magnetisation does not influence our gravitation/temperature effect. 
Nos. 3 and J.—These are taken last, their effect being the most difficult of all 
errors to dissect out from the net result. No calculation can be made of the forces 
set up due to their action since we do not know the amount or distribution of the 
supposed irregular heating on the inner face of the vacuum tube. The apparatus 
was designed and fitted up to avoid these errors to the utmost. There are a great 
number of layers of cotton wool, paper and flannel both inside and outside the helical 
water-jackets. Also extra screens were in some cases arranged to keep heat from the 
vacuum tube. The water passing through the water-jacket was steady in temperature 
for long intervals ; any small rises in temperature due to the source would, in a long series 
of experiments, pair off against similar falls. The water flowed at a great rate, say, 3 
to 6 litres/minute. It is hard to believe that any heat from the thickly-lagged lead 
spheres could penetrate the water-jacket and cotton wool under these conditions. 
But one can never be sure, without special tests, that heat will not find some joint 
or weak spot in the lagging and so reach the vacuum tube somewhere. If the inside 
of the vacuum tube were warmed irregularly we might have radiometric pressure 
effects ; but even then one would expect no change in the range (see Table II., 
column 7), but rather a removal of the same range up or down the scale. The special 
tests applied were : — 
(a) The tap water was heated before entering the helix "and it was found that a 
change in telescope readings occurred at first, but that when the temperature became 
Steady, though raised, the readings became normal. 
( b ) More and yet more wrappings were put over the helixes, so that weak places, if 
any existed, would be covered up. It will be observed that the numbers in column 2, 
