It is well known that displacements of spectrum lines during a photographic exposure can be caused by changes in the temperature of the dispersing apparatus and of the air in the spectrograph. In the case of grating instruments elaborate precautions are sometimes taken to keep the room in which the spectrograph is housed at a constant temperature. Birge* has described the precautions taken at the University of Wisconsin, where the grating spectrograph is mounted in a double-walled room, the space between the walls being kept constant by suitable electric heating. In the photography of faint spectra prismatic spectrographs are sometimes to be preferred to gratings on account of their superior light-gathering power, and it is of importance to consider the errors in measurement and loss of definition due to variations in temperature. The change in refractive index with temperature appears to vary rather considerably with different melts of glass. Giffordf has measured the temperature coefficients for a series of glasses in air, and his results show that these are positive for all the glasses investigated, with one exception, Fluor Crown, which would not be suitable for the construction of prisms for spectrographs. The change in refractive index per degree Centigrade for glasses of the type suitable for the construction of prisms is shown to be of the order of 5 X 10
-6
, and since for a typical glass at λ = 5875 A, the change per Ångstrom unit is 9 X 10
-6
, the shift in wavelength per degree Centigrade would be upwards of half an Ångström unit. The temperature coefficient of some glasses seems to be considerably greater than the above value. Displacements of spectrum lines are also effected by changes in the barometric pressure during long exposures. There are, of course, many days on which this effect may be neglected, but this source of error does not appear to have received the attention that it merits. The refractive index of air at 760 mm. pressure and 20° C. is about 1•000275, so that if we assume that
n
p
=
n
760
— [(
n
— 1) (760 — p)/760] the change in n per cm. pressure is about 3·6 X 10
-6
, and taking the change in
n
per Ångström unit at D
3
as 9 X 10
-6
, we expect a shift of about 0·4 A per cm. pressure. A change in pressure as great as 1 cm. during the course of an exposure would be unusual,but changes exceeding 1 mm. are not infrequent during long exposures, and it is evident that they cannot be ignored. It is usual to expose the comparison spectrum, comprising the lines of standard wavelength, either simultaneously with the spectrum under investigation, or in instalments at regular intervals, but it has been shown that errors due to displacements* can only be eliminated in this way in the case of lines which are symmetrical, and in any case there is a loss of definition and resolving power with a loss of accuracy of the settings in the course of measurement. The experiments described in this communication relate to a prism spectrograph recently constructed for the writer by Messrs. Adam Hilger, and although the methods adopted and the conclusions reached seem fairly obvious, they may be worthy of record in view of the fact that satisfactory results were obtained only after trials extending over many months, and the details of construction appear to be of vital importance.
XIX. On the photographic arc spectrum of iron meteorites
