24 KENNETH HARTLEY 
measured. The ideal material for the spring would be a nonmagnetic metal 
with a high elastic limit but a low value for Young’s modulus and a low tem- 
perature coefficient. This combination might be found in one of the less com- 
mon metals or alloys but information on the elastic properties of these 
metals, particularly temperature effects, was not to be had. Assuming that 
the temperature coefficient of elasticity would probably be smaller in a metal 
with very high melting point led to the consideration of tantalum, molyb- 
denum, etc., and several springs were made of approximately the dimensions 
that would be required and tested in that form. The final choice was an alloy 
of tantalum and tungsten which seems to be fairly satisfactory. The spring is 
wound with a high initial tension so that when carrying its full load the coils 
are separated only about half a millimeter, making the total length of the 
stretched spring less than ten centimeters. The frame of the instrument is 
made of an aluminum alloy, which has the highest coefficient of expansion of 
any suitable metal, and is constructed on the principle of the old grid-iron 
compensated pendulum so that the total length of the high expansion 
members is about 80 centimeters, the low expansion rods are invar steel. This 
compensation is mainly for the purpose of avoiding too great a change in the 
relative position of the parts when the working temperature is altered. In 
order that the whole instrument be in a condition of thermal equilibrium and 
free from temperature stresses it must be kept at a constant temperature day 
and night during a whole series of observations; it is therefore enclosed ina 
new type of portable thermostat, designed on the principle worked out at 
the Bureau of Standards two or three years ago, which controls the tempera- 
ture to within one ten-thousandth of a degree. The heating is electrical and 
is designed to work with a six volt storage battery. 
A study of the action of this instrument while the temperature is either 
rising or falling will quickly convince any one that any plan for making a 
temperature correction, instead of maintaining a constant temperature, is 
wholly out of the question. In the first place, even with the best possible com- 
pensation the temperature would have to be measured to within one hun- 
dredth of a degree, and in the second place, it is impossible to have the entire 
apparatus at the same temperature unless it is kept constant for a consider- 
able time. In the present instrument if the temperature is altered by two or 
three degrees it requires at least five hours to reach a condition of equilibrium 
and if the change is much greater it may take all day, but with thermostat 
control the temperature effects are also brought within one part in ten million. 
The change in density of the air, due to varying barometric pressure, may 
alter the buoyant force on the suspended mass by several times the quantity 
that we are attempting to measure, so the case is closed air-tight and the two 
operating rods are surrounded by a new type of mercury seal which prevents 
the passage of any air in or out even with a difference of pressure of more than 
two inches of mercury. 
The effects of elastic hysteresis are eliminated by locking the weight and 
spring in the zero position between observations, so that the tension on the 
spring is never varied by more than the difference between the force of gravity 
168 
