OS SOLAR RADIATIOxV. 41 



at different distances and different rates of heating, that the equations 

 already given did not satisfactorily represent the observations. In con- 

 sidering the possible sources of constant error inherent in the method, it 

 seemed unlikely that the assumption that the rate of loss of heat was 

 proportional to the difference of temperature (q constant) could be 

 seriously in error, as the whole difference of temperature did not exceed 

 one or two degrees. The most probable explanation of the discrepancy 

 appeared to be that time was required for the uniform distribution of 

 heat through the disc, and that the indications of the thermo-junction 

 were retarded by conduction of heat along the wires, and by lag in the 

 movement of the galvanometer coil, which was necessarily very dead-beat 

 when shoi't-circuited on the couple. These various sources of error could 

 all be approximately represented by assuming a constant time-lag in the 

 readings ; a type of error which was necessarily inherent in the method, 

 and could not have been detected by the experiments recorded in 1896. 

 In order to eliminate the time-lag from the equations, it is only necessary 

 to take two observations in addition to the final steady temperature. If 

 we write the equations in the form (3) already given, and take the differ- 

 ence, we thus obtain 



q{t"-t')=2-3026\og^,{6°-6')/{d°-6") . . (4) 



Treating the observations already given in this manner we find 



q=-OlUO. r=-01232. I=-00.518. Time-lag=6-15 sees. 



With only three observations it is of course always possible to calculate 

 a value of the lag to satisfy the readings exactly, but it appeared that a 

 .similar assumption satisfied the observations withui the probable limits 

 of error in those cases also in which a larger number of readings were 

 taken. 



Defects of the Copper-cube Actiiiometer. — The excessive value of the 

 time-lag observed in the observations with this apparatus appeared to be 

 partly due to the impossibility of securing uniform illumination of the 

 disc by the aperture method. It was necessary that the disc should be 

 large enough to catch the whole of the radiation passing througli the 

 aperture in the cube, and this could not l)e secured without leaving a con- 

 siderable margin at the edge of the disc which was either not illuminated 

 at all, or only partly illuminated by the penumbra of the aperture. With 

 the lamp at 00 cms. it was necessary to use a disc 1"40 cm. in diameter 

 for an aperture of I'OO cm. diameter. This was the more necessary because 

 the construction of the apparatus, and the method of screwing in the copper 

 plug by which the disc was supported, made it extremely difficult to centre 

 the disc accurately, and to direct it so as to receive the rays normally and 

 centrally. Another serious defect to which allusion has already been 

 made, was the variation of temperature of the cube and the copper plug, 

 which although greatly reduced was not entirely eliminated by the cotton- 

 wool wrappings. For these and other reasons it was decided to design a 

 new form of actinometer for the application of the same method in a 

 manner more convenient for laboratory use. 



Tuhe-form of Radio-calorimeter. — The terms ' actinometer,' ' bolo- 

 meter,' and ' radio-micrometer,' which are otherwise suitable for instru- 

 ments of this class, have acquired special significations, and are in general 

 use for instruments which are not designed for absolute measurements, 



