ApriL 28, 1923] 
_ cancel one another. The light-pattern viewed by the | 
, eyepiece is thus a line of alternate bright and dark 
: fringes. At first sight it might seem that the fringes 
. would continue of equal brightness to a great distance ; 
but actually they soon fade away, because in the more 
oblique directions there is interference, not merely 
between the two apertures, but between different 
parts of the same aperture. In fact the fringes appear 
as fine detail in the midst of the small diffraction disc 
which would be formed by either of the apertures singly. 
Arrange that the two apertures are movable, and 
widen their distance apart. The points Q,, Qo, Q, now 
come closer together, that is to say, the fringes con- 
tract. Decrease the distance and the fringes spread 
out. It is a simple matter to find a formula giving 
the width of the fringes for any given separation of 
the apertures. 
When the object viewed is a double star—two points 
_ of light—each point will produce its own line of fringes, 
and these will be superposed if the double star is a 
close one. Jt may happen that the two systems of 
fringes are in step, in which case the alternate bright 
and dark spaces will be conspicuous: but if they are 
at all out of step the pattern will be blurred. Remem- 
bering that we can alter at will the width of the fringes 
by varying the separation of the apertures, we can 
adjust them, so that the bright fringes for one com- 
ponent coincide with the dark spaces for the other 
component. If the two systems are of equal bright- 
ness one system will fill the gaps in the other, leaving 
merely a line of uniform brightness ; if the two com- 
ponents are of unequal brightness the same adjustment 
will give minimum visibility of the alternation of light 
and shade. Varying the distance between the aper- 
tures this critical position can be fixed with consider- 
able precision ; and for close double stars of not too 
unequal magnitude the separation is measured in this 
way much more accurately than with a micrometer. 
The disc of a single star or planet is likewise measured 
by finding a position of the apertures for which the 
fringes disappear. It is not now a problem of two 
points producing overlapping fringes, but every point 
of the circular disc produces a fringe-system, and the 
effects must be summed. When the diameter of the 
disc is 1-22 times the width of the fringes (z.e. the dis- 
tance from one bright fringe to the next) the integrated 
effect is uniform illumination and the fringes disappear 
altogether. It is not much use trying to see in one’s 
head a result which is more fittingly the subject of 
algebraic calculation ; but we may notice in a general 
way that with this ratio the two outer quarters of the 
disc fall at places where the central half is producing 
dark spaces. That indicates roughly how the different 
portions of the disc compensate one another. The 
observation consists in varying the separation of the 
aperture until the fringes disappear ; the diameter of 
the disc is then 1-2 times the fringe-width calculated 
for that separation. 
The possibilities of a method of this kind had been 
explored to some extent before Michelson took up the 
problem ; indeed Stéphan in 1874 had attempted un- 
successfully to detect the discs of stars by this means. 
We owe to Michelson the practical demonstration of its 
success. His first paper appeared more than thirty 
years ago in the Philosophical Magazine for July 1890. 
NO. 2791, VOL. 111] 






























NATURE 


afd 
The next year he followed up the theory by measuring 
the diameters of the four satellites of Jupiter at the 
Lick Observatory. The method proved entirely 
successful, and his measures were afterwards closely 
confirmed by Hamy at the Paris Observatory in 1899 
using the same device. The great value of the method 
seemed to be proved ; it was thoroughly tested ; and 
it forthwith lapsed into oblivion. 
In 1919 Michelson again took up the matter with 
energy. He made observations in August with the 
4o-in. refractor at Yerkes which were found to be 
encouraging ; and he went on to use first the 6o-in. 
and then the 1oo-in. at Mount Wilson with the more 
ambitious design of surpassing the highest resolving 
power yet reached. At Mount Wilson he had the 
co-operation first of J. A. Anderson and afterwards of 
F. G. Pease. A great success was quickly obtained 
with the double star Capella. Capella is a spectro- 
scopic binary with a period of 104 days. It was 
known that the distance of its components must nearly 
approach the limit of visual detection, but attempts 
to observe it visually had failed. (I may remark that 
that is a rather controversial statement to make— 
particularly at the R.A.S.—but the controversy is now 
ancient history.) With two narrow apertures in the 
beam from the too-in. mirror the fringes were ob- 
served and then brought to minimum visibility by 
varying the position angle and separation of the aper- 
tures. The changing position angle and distance were 
traced through the revolution. Anderson’s measures, 
afterwards continued by Merrill, have given a very 
accurate orbit. The separation of the two components 
varies from 0-04” to 0-05". From a comparison of this 
visual orbit with the spectroscopic orbit we find the 
parallax of Capella and also the mass. The parallax 
is 0-063”, and the components are respectively 4:2 and 
3°3 times as massive as the sun. The parallax does 
not differ much from that given by trigonometrical 
and spectroscopic determinations; but these were 
very rough values, whereas the interferometer parallax 
is presumably of the highest order of refinement. I 
suppose that the mass determinations are about the 
best we have for any star. But what is especially 
important is that Capella is the only giant star 
for which we know both the mass and the absolute 
luminosity. 
I may perhaps be allowed to refer to a personal 
interest in this first big result of Michelson’s method. 
Capella now supplies the chief lacking constant in the 
radiative theory of stellar equilibrium, for which I 
had waited five years. It is, I think, generally con- 
ceded that the absolute magnitude of a giant star 
mainly depends on its mass, and theoretical formule 
can be found expressing the law of dependence. But 
we need to know one pair of corresponding values in 
order, as it were, to anchor the formule. Hitherto 
that correspondence could only be guessed roughly 
from statistical knowledge of the average luminosity of 
giant stars and an estimate of the corresponding 
average mass based on our general knowledge of the 
masses of stars (which, unfortunately, relates chiefly 
to dwarf stars). Having now the exact figures for 
Capella, we can substitute a precise determination 


2 The masses of a Centauri, Sirius, and Procyon may have about the 
same accuracy, but I do not think that any others reach this standard, 
