EXPERIMENTAL DETERMINATION OF WAVELENGTHS IN THE INVISIBLE 



PRISMATIC SPECTRUM. 



A COMMUNICATION TO THE NATIONAL ACADEMY OF SCIENCES, IN AFRIL. 1K83. 



[Note. — The following investigation w.as made at the expense of the Bache fund and is i)nl)lisbed here by the 

 permission of its trustees.] 



In September, 1881, while eugaged upon Mount Whitney in measnj ng- with a linear bolometer 

 the heat in the invisible spectrum of a flint prism, I came upon a hithe. lo unknown cold band whose 

 deviation indicated a (probably) very great wave-length.* We have had up to the present time 

 no way of measuring such wave-lengths directly, but are accustomed to determine them by more 

 or less trustworthy extrapolation formnlre, the best known of which is Cauchy's. Accordingly, I 

 attempted to calculate the wave-length by Cauchy's formula, but was conducted to an impossible 

 I'esult — the formula declaring that no such index of refraction as I had measured was possible in 

 the prism in question. But the measurement was a fact beyond dispute, and this drew my attention 

 to the grossness of the errors to which the customary formulie may lead. 



Every prism gives a diftei-ent map of the spectrum, nor when we find a band or line by the 

 prism have we any means of fixing the absolute iilace, except by a reference to the normal or wave- 

 length scale, or to one derived from it. 



It is desirable to define at the outset the sense in which the term " normal " is here used, as a 

 synonym for " wavelength" spectrum. 



The amount of energy in any region of the spectrum, such as that in any color, or between any 

 two specified limits, is a definite quantity, fixed by facts which are independent of our choice, such 

 as the nature of the radiant body, or the absorption which the ray has undergone. Beyond this, 

 nature has no law which must govern us in representing the distribution of the energy, and all 

 maps and charts of it are conventions. 



If the length of the spectra formed by any two different agents, such as a prism and a grating, 

 be made equal, it does not then follow that the lengths of similar portions must be equal. In the 

 case supposed, we observe in fact that the red portion (for instance) of the prismatic spectrum will 

 be narrower than the red portion of the second. But since the amount of energy in the red must 

 be really ^he same in both, we must, in a graphic representation of this energy, increase the height 

 of the ordinates in the red of the prismatic spectrum, so that the areas shall remain the same. 



The position of the maximum ordinate is then (in one sense), a matter of choice, and fixed 

 only by the scale we elect to emjdoy. We find, for instance, in the prismatic spectrum, that this 

 ordinate is in some part of the infra-red, depending on the particular prism used; while, in the 

 grating spectrum, it is, under the same circumstances, always in one part of the yellow; and we 

 might conceive of an apparatus which should always exhibit it in the ultra-violet, or which should 

 even show the same energy at one wave-length as at any other, or embody any other arbitrary 



* Since designated as " C ." 



149 



