Septeiviber 13, 1901.] 



SCIENCE 



409 



passed through the pi-ism at minimum devia- 

 tion. The rotation of the pi-ism corresponding 

 to the entire range of wave-lengths here studied 

 is less than 1° with the rock-salt prism — and 

 hence it is necessary to be able to determine 

 the angular position of the prism with rather 

 unusual accuracy if the position of the bolo- 

 meter in the spectrum {i. e., the wave-length) is 

 to be accurately known. The aim has been to 

 have the maximum error in a single determina- 

 tion of the position of the bolometer in the 

 spectrum not greater than 0^^.6, corresponding 

 in the usual case to 0.1 mm. on the photo- 

 graphic plate. The critical part of the mechan- 

 ical connection between the spectrometer and 

 plate is a worm and wheel-segment by Warner 

 and Swasey, which is so good that the above 

 rigorous conditions can probably be nearly sat- 

 isfied. When in use the spectrometer and plate 

 are moved uniformly by clockwork which is, 

 however, quite independent of the accurate 

 mechanical connection between them, so that 

 its errors do not enter. The photographic en- 

 ergy curve thus obtained by the motion of the 

 spectrum past the bolometer strip will show 

 besides the depressions due to the absorption 

 Hues sought, many other depressions and ir- 

 regularities due to disturbances of a mechanical, 

 electrical, magnetic or thermal nature, and a 

 large part of the work has been an attempt to 

 reduce to a minimum these accidental disturb- 

 ances in order that the depressions in the curve 

 due to real lines might be more clearly identi- 

 fi.ed. It is impossible to consider here the causes 

 of these irregularities — some coming from the 

 bolometer circuit, some from the galvanometer, 

 some from the battery, or the steps which have 

 led to the gradual reduction of them, though 

 these fill what is in some ways the most inter- 

 esting chapter. Suffice it to say that careful 

 comparison of a number of the best curves, or 

 holographs as they are called, has resulted in 

 the more or less certain identification of over 

 700 lines in the region from ^ = 0.76/^ to a = 5.3« 

 whose wave-lengths are determined with a mean 

 probable error of two or three Angstrom units, 

 the results being given in tables, containing also 

 indications as to the character and grouping of 

 the lines. A graphic representation is also 

 given in the form of line spectrum maps, both 



normal and prismatic, either di-awn by hand 

 from the tabulated results, or transformed in a 

 semi-automatic way from some of the curves 

 themselves. This process seems of doubtful 

 value, however, as the line-spectrum conveys 

 in reality less information than the curves them- 

 selves, while it is apt to produce a false impres- 

 sion of authenticity. Curves and tables are also 

 given illustrating the changes which occur, 

 seasonal and otherwise, in the intensity of 

 many infra red absorption lines ; changes whose 

 existence is here definitely fixed, but which 

 will have to be further investigated before their 

 cause can be assigned. 



The most important subsidiary research has 

 been the determination of the dispersion curves 

 of rock-salt and fluorite ; the former by direct 

 comparison (so to speak) with a grating, the 

 latter by comparison of fluorite holographs with 

 rock-salt holographs of the solar spectrum, and 

 measurements of the deviation of the same lines 

 in both. Langley's early method was substan- 

 tially repeated here in the work with rock-salt, 

 and is briefly as follows : 



Radiation from whatever source is used falls 

 first upon the slit of a Rowland concave grating 

 spectroscope, the eyepiece of which is replaced 

 by a second slit ; the radiation passing this 

 second slit, from the grating, goes through the 

 optical train of the prism spectroscope and is 

 brought to focus at the proper place in the dis- 

 persion spectrum of the particular prism used. 

 If the grating is set so as to let through the 

 second slit light of a known (visible) wave- 

 length in (say) the 3d order spectrum, waves of 

 I and 3 times this length will also pass through, 

 belouging to the 2d and 1st order spectra. 

 Hence in the dispersion spectrum of the prism 

 there will be three distributions of energy, 

 more or less sharp according to the widths of 

 the two slits ; and if a holograph of this spec- 

 trum is taken in the usual way, three maxima 

 will appear, separated by regions indicating no 

 radiation. If on the same plate a record of the 

 ' A ' line is made, the relative deviation of the 

 maxima (due to known wave-lengths) with re- 

 spect to the ' A ' line, can be determfned by 

 measurement of the plates. Such records, 

 combined with the determination of the abso- 

 lute deviation of the ' A ' line, and of the angle 



