2 Prof. A. Schuster on a Sample 
A,, and so on as the optical distances of the reflecting strips 
gradually increase from A, to A,. Hence the disturbance at 
F consists of a series of impulses, 
following each other at intervals equal 
to the period of a homogeneous wave 
which, starting from the same lumi- 
nous point, and reflected by the 
grating, would have its first principal 
maximum at F. All this, of course, 
is simply the elementary illustration, 
first given, believe, by Lord Rayleigh, 
of the ordinary action of a grating 
when analysing white light. The  s <a 
question now is: How can the im- 
pulses which succeed each other at 
F be made to interfere? Clearly only 
by retarding those which first reach 
F, or accelerating those which reach 
| 
= | 
| 
J 

that point last. A plate of appro- 
priate thickness introduced from the 
left-hand side of the figure as it 
is drawn, can be made to answer 
the purpose. If, on the contrary, the same plate were 
introduced on the right-hand side, it would only retard those 
impuises which already arrive late, and therefore no inter- 
ference could take place. This is really all that need be 
said in explanation of the bands; but a more detaiied consi- 
deration of this view of the problem leads easily to a clearer 
expression for the calculation of the best thickness of the 
interposed plate than the more elaborate calculations of 
previous investigations. 
The best thickness is secured when the whole series of 
impulses is divided into two equal portions, the impulses. 
arriving in pairs simultaneously at F. If A, be the central 
line of the grating, the retardation ought to be such that the 
impulses coming from A, and from A, reach F at the same 
time. If N be the total number of lines of the grating, the 
best retardation is therefore $NA, and the plate should be 
pushed sufficiently far into the beam to affect halfits width. 
The wave-length \ here means the wave-length of that homo- 
geneous train of waves which has its first principal maximum 
at F’, so that the retardation of each impulse compared with the 
nextis A. If the retardation is either greater or smaller, some 
of the impulses arrive too soon or too late to overlap others, 
and the bands are less clear. If the retardation has more 
than twice its best value, the series of impulses from A, to 
FE 
