SOLAR CONSTANT OF RADIATION"— ABBOT. 325 



strument. Besides this, there is reason to believe that the instrument 

 reads too low, even at the first. 



In most pyrheliometers, as in that of Pouillet, there is a blackened 

 exterior surface for the reception of the solar radiation, behind 

 which lies a device for measuring temperature. There are two pos- 

 sible paths for the heat produced, one by conduction back to the tem- 

 perature device, one by convection and radiation forward into the 

 air. This second part of the heat is lost, and is undetermined, though 

 not large in amount. Besides this loss is a second by direct reflec- 

 tion of rays from the surface. This second loss is usually allowed 

 for, but its determination is not easy. Both these sources of error are 

 avoided in the hollow chamber instrument of Michelson. 



When in 1902 we began at Washington the study of the " solar con- 

 stant " we fortunately, though quite innocently, did not employ 

 Angstrom's pyrheliometer. We had received from the late M. Crova 

 two of his alcohol actinometers, and of these he said in a letter, with 

 delightful naivete, that they were good secondary instruments and 

 only required to be calibrated by comparison with any satisfactory 

 standard. At that time there was no standard. So we cast about for 

 one, and, following Tyndall, who had followed Pouillet, I had our in- 

 strument maker, Mr. Kramer, prepare a shallow, circular copper box 

 with a thermometer inserted at the side, and with mercury filling 

 the box to the blackened cover. The whole was surrounded by a 

 wooden chamber to keep off the wind. By calorimetric measure- 

 ments we attempted to get the water equivalent of this mercury 

 pyrheliometer, and our subsequent measurements were all given, 

 for years, in terms of the scale it furnished. We soon found our 

 mercury pyrheliometer more convenient than the Crova actinometer, 

 and abandoned the use of the latter. Afterwards we recognized 

 that, on account of the great variation of the specific heat of alcohol 

 with change of temperature, we should have been all at sea if we 

 had continued to use Crova's instrument. Later we dispensed al- 

 most wholly with the mercury and used a solid, circular, thin 

 copper block, with a radial hole for inserting the cylindrical bulb 

 thermometer, using only enough mercury to make good heat conduc- 

 tion to the thermometer. Finally we have bent the stem of the ther- 

 mometer at right angles, employed a steel-lined silver block, have 

 equipped the instrument with various effective little auxiliary devices, 

 and by the aid of a grant from the Hodgkins fund by the Secretary 

 of the Smithsonian Institution, have sent several of our silver-disk 

 pyrheliometers to Europe, to promote international agreement in 

 pyrheliometry. 



But although our scale of pyrheliometry was fortified by com- 

 parison of numerous copper and silver disk instruments among them- 



