14 



NATURE 



[November i, 1894 



that during this same second the plate will have been 

 lifted automatically through one-sixtieth of a centimetre 

 of space. 



This is one relationship of time and space in actual use 

 here, though others may be established by the use of the 

 change-wheels with which the apparatus is provided. 

 The essential thing is that the plate shall show with great 

 precision, and even on simple inspection, not only the 

 inflections of the energy-curve there written down, but the 

 exact relative position in the distant spectrum which the 

 bolometer thread occupied at the moment it caused the 

 disturbance. In the case assumed, for instance, if we 

 suppose that the record on the plate commences with the 

 part of the spectrum whose angular value is 40 , then, 

 since : millimetre corresponds to 6 seconds of arc, and 

 so on, the e.xistence of an inflection on the plate at 30 

 centimetres, 3 millimetres and seven-tenths of a milli- 

 metre, would show that the disturbances originated at 

 the point in the spectrum corresponding to an angular 

 measure of 40" 30' 22 'a. 



If the arm which carries the bolometer is n metres 



long, and if the thread of the bolometer is - metres in 



diameter, the angular value of the bolometer thread is 



— . At present the linear width of the bolometer thread 

 mn 



is not very materially less than formerly, but it is used with 

 a longer arm, and its virtual width is accordingly less. In 

 present actual practice (to use round figures) the optical 

 arm carrying the spectrum across the bolometer, is five 

 metres in length ; and if the bolometer thread be one- 

 twentieth of a millimetre in width, its angular value is 

 evidently jsoVuis o*^ 'he radius of the circle in which it 

 moves, or a little over two seconds of arc. AVhen the 

 heat is distributed over so large an area, that part of it 

 which falls on a thread of given diameter is, of course, 

 proportionately less, so that the greater precision of 

 measurement demands a more sensitive construction of 

 the bolometer, as well as a more accurate mechanism for 

 pointing it. Improvements have accordingly been in- 

 troduced in the construction of the bolometer, and a 

 need for greater sensitiveness in the galvanometer has 

 necessarily gone with them. This increased sensitive- 

 ness has caused increased liability in the latter to both 

 systematic and accidental perturbations, and the elimina- 

 tion of these has been found the most formidable diffi- 

 culty of the whole process. It has been effected, largely, 

 by placing the whole apparatus under constant tempera- 

 ture conditions. 



I take pleasure also in acknowledging the advantage I 

 have found in using both Prof. Hoys's quartz threads and 

 the extremely small mirrors which he, 1 think, first advo- 

 cated in connection with the well-known form of galvano- 

 meter due to Lord Kelvin. These and other collective 

 improvements made in the bolometer and in the galvano- 

 meter, have now made the former sensitive to changes of 

 temperature in its strip which are demonstrably less than 

 I 1,000,000 of a degree Centigrade. 



These are the principal pieces of apparatus, though I 

 should mention that a method has been found by which 

 the very large salt prisms used can be preserved in per- 

 fect polish while exposed to all the usual casualties of 

 observation. The actual prism in most frequent use was 

 made from a block of salt exhibited at the World's F'air 

 by the Russian Government, and presented to the 

 .Smithsonian Institution by its Commissioners. It is 

 about eighteen centimetres, or over seven I'.nglish inches, 

 in height. 



Before entering upon a description of the results 

 obtained, I desire permission to speak of the aid I have 

 received from the gentlemen whose assistance I have 

 been fortunate in securing; first, to Dr. Ilallock, then to 

 Prof. Hulchins, Mr. Hubbard, and Mr. C. T. Child, and 



NO. 1305, VOL. 5IJ 



lately to Mr. F. L. O. Wadsworth and Mr. R. C. Child J 

 the imprint of the labours of the two latter gentlemen 

 being upon almost all the details of the more recent 

 work. 



Results. 



Let us recall that the infra-red spectrum from a rock- 

 salt prism, such as that used here, is extremely con- 

 tracted as compared with one from flint, and still more 

 contracted as compared with the wave-length scale. The 

 portion of the spectrum presented by photography 

 reaches a little below the band whose wave-length is 

 about i/x, and this was asserted by one of the most 

 eminent living authorities on the subject (Dr. John W. 

 Draper), when the writer commenced this work fifteen 

 years ago, to be the absolute end of the heat spectrum. 

 The writer has, however, since carried his investigations 

 by direct measurement to five or six tmies this wave- 

 length, and by indirect measurement much farther still, 

 though what is here now exhibited does not go beyond 

 a wave-length of about 4/i. The invisible heat spectrum 

 of a 60" rock-salt prism through this great wave-length, 

 includes only somewhat less than two degrees of arc, and 

 the first of these degrees contains the greater proportion 

 of the energy. 



On referring to the illustrations exhibited to the 

 Association in 1882, or even to later publications of 

 results obtained by rock-salt prisms, though with the 

 old method, it will be seen that there are shown in the 

 latter publication about a dozen measured inflections of 

 the energy curve below i/i'5, and it may be remembered 

 that this curve was obtained only by two years' assiduous 

 labour. 



We have now before us three energy curves obtained 

 by the new method, each exhibiting the whole infra-red 

 spectrum under examination, with about a hundred 

 inflections. These curves are nearly, but not exactly, 

 similar. 



The three were obtained in the same day, each from 

 an entirely independent observation, so that each has 

 given, in a fraction of a day, many times the results 

 previously obtained by two years of labour, and, as it will 

 be later shown, has given these results with a notable 

 gain of accuracy. 



But this is not all. These three curves have been 

 taken with a rapid movement of the clockwork and a 

 brief swing of the galvanometer, so as intentionally to 

 suppress all minor inflections and to introduce only the 

 leading features of the spectrum, as shown in eighty or a 

 hundred of the leading inflections (lines), or groups. 



This new bolometric method has, however, as will be 

 shown later, a capacity of resolving these into nearly 

 twenty times that number, the minor inflections having 

 been thus designedly suppressed here at first, to better 

 show the character and position of the principal ones. All 

 these energy spectra, by the new as by the old method, 

 are, of course, subject to the slight changes due to invisible 

 clouds constantly passing before the sun, which, with 

 the change of the sun's altitude, and of the consequent 

 lengthening path of its rays, prevent any one of them 

 from being exactly like the other ; while, at the same 

 time, everyone here may satisfy himself by direct 

 inspection of the results before him, that there is scarcely 

 any single one of their inflections which is not reproduced 

 in the other two, in exactly the same place, though 

 probably not exactly in the same degree ; and when we take 

 different spectral traces, made at different hours of the 

 day, and even on diflerent days of the month — traces 

 which are absolutely independent of each other — and 

 superpose them, experience shows that we may expect to- 

 see such an agreement as that in the three here chosen 

 at random for illustration, or in the more detailed one. 

 where the relative probable error is less than one second 

 of arc. Three such traces only are here given (to prevent 

 confusion), but if we follow these coincidences through 



