Vol. 6, 1920 
PHYSICS: C. BARUS 
501 
apprehended from viscosity may be estimated from the following observa- 
tions, figure 4, of the yield of the wire ( AA^", vertical) with the gram excess 
at the end of the balance beam. The graphs shown are for the 4th to the 
7th twists, alternately in opposite directions. The others, eleven in all, 
were similar. The curves show that if measurement is made within 3 
minutes after twisting, the error from viscosity, if ignored, would not ex- 
ceed 8g/g = 3 X 10 If, however, the coefficient of viscosity is known 
and allowed for, the error of 3 X 10~^ is improbable. 
If the graphs be used to compute the absolute viscosity of the metal 
for the given twist and diameter, it is found to increase in 100 minutes 
from about 4 X 10^^ to 3 X 10^^ and thereafter is nearly constant. The 
twist was about 1.4° per linear centimeter; but in alternating, the effect of 
this is about doubled. 
Observations on the Permanently Twisted Wire. — The readings were made 
two or three times daily, as a rule, for over a month. They were ade- 
quately recorded in graphs like figure 5, for example, figure 6 being the 
corresponding temperature graph. As the twist produced by the excess 
of weight of about 1 gram was here also about 25°, we have bg/g = 1.6 X 
10"^ for each of the divisions (10 ~^ cm.) of the scale. The laboratory 
temperature was very variable from day to day; and the effect of this 
on the graph is at first sight disconcerting, though it contributes essentially 
to the interpretation. In figure 5 the initial mean rate of yield referable 
to viscosity, may be estimated as = 0.0009 cm. per day. This 
makes bg/g vary 0.00014 per day, a quantity in itself too large to be used 
as a correction, with confidence; after a month, however, this nearly 
vanishes, but a much more serious consideration, the thermal variation 
of rigidity, remains. It is this feature which makes the graph so exceed- 
ingly jagged. If we take the large drop between the 8th and the 10th 
day to estimate this effect, the data would be: / = 22°, 15°, 20°; 10^ AN = 
8.1, 2.6, 8.7 cm., which is equivalent to AN/ At = 10~^ cm. per degree; 
or 8g/g = 2 X 10 nearly, for each degree of fall of temperature pro- 
moting increased rigidity. Excellent observations on the torsional rigidity 
of iron and steel are given by Pisati (Landolt and Boernstein's tables). 
The coefficients are 2.1 X 10~^ for iron, and 2.3 X 10""^ for steel, therefore, 
even larger than the graph indicates. In fact, the present interferometer 
method could easily be modified to measure this constant accurately. 
Summary. — In the above experiments as a whole, there is thus very 
little promise of arriving at the variations of g, with precision, by using 
torsional apparatus, so long as there is any marked change in the tempera- 
ture of the environment. The discrepancy of the method is to be re- 
ferred less to the viscosity of the metal and its thermal coefficient, than 
to the decrease of rigidity with rising temperature. 
There is, however, one possible way out of the difficulty and this points 
