156 Prof. S. P. Langley on 



is accurately self-centred in the cylinder. For protection 

 from air-currents, since the obscure heat studied will be 

 stopped by a glass cover, we must make use of the special 

 device I have described in the memoir just referred to, of 

 successive chambers or drums separated by diaphragms with 

 a common central aperture. 



With these precautions, and with the special adjuncts 

 before described, a bolometer with a strip ^ millim. wide 

 can be set by the invisible heat alone to within 10" of arc, 

 while in the ordinary use of the linear thermopile we are 

 liable to errors of a considerable fraction of a degree. Even 

 with a bolometer 1 millim. wide, it will be subsequently seen, 

 we can set to one minute of arc. This refers only to the 

 precision of pointing attainable ; we will consider the sensi- 

 tiveness of the instrument later in connection with the 

 galvanometer. 



GrALVANOMETEK. 



It must be remembered that while the nominal sensitive- 

 ness of a galvanometer can be increased to any extent by 

 increasing the astaticism of the needle (quite as nominal 

 power can be multiplied to any extent on a telescope by 

 altering lenses at the eyepiece), that the real or working 

 capacity depends upon the ability to always obtain a like 

 result under given conditions. Accordingly we have con- 

 tinued to devote great pains to extend our original concep- 

 tion, so as to make the galvanometer, as well as the bolometer, 

 not merely an indicator of heat, but a real " meter/' which 

 shall distinctly answer the question "how much ?" as to almost 

 infinitely minute amounts of energy. 



For the benefit of any physicists who may desire to repeat 

 these experiments, we may observe that we have found the 

 bolometer capable of almost unlimited delicacy of perception 

 of heat, but that our chief trouble has arisen from the diffi- 

 culty of constructing a galvanometer suitable to develop its 

 full capacity for exact measurement. We have been unable 

 to find among galvanometers ordinarily constructed one 

 capable of indicating with precision changes in the amount 

 of current of much less than 1)00 , J)Ul0 of an ampere. It was in 

 the construction of a galvanometer designed to measure the 

 heat in the spectrum of the moon, that we acquired the ex- 

 perience which we have utilized in the present researches. 



A reflecting-galvanometer of the form devised by Sir 

 William Thomson has been used for the basis of our construc- 

 tion and altered as follows. (For several of the changes 

 described I am indebted for suggestions due to the great 



