452 PHYSICS: W. N. BIRCHB Y PROC. N. A. S.

The divergences in a are certainly no larger than the different errors of

measurement. So far as I see, therefore, it does not seem possible to reach
the relatively low value (a = .81) usually quoted for the diffusion of hydro-
gen into air by the present method.
* Advance note from a Report to the Carnegie Inst. of Washington, D. C.
** These PROCEEDINGS, 10, 1924, P. 153.

WHITE LIGHT INTERFERENCE FRINGES WITH A THICK

GLASS PLATE IN ONE PATH*
BY W. N. BIRCIBY
CAiIPOR-t INsTiTuT or TuCENOLoGY
Communicated, September 10, 1924
The object of this paper is to describe and account for certain interfer-
ence phenomena which are observed with white light, when a thick plate of
glass, or other dispersive medium, is placed in the path of one of the beams
in the Michelson interferometer. The mirrors are adjusted in sodium light
for circular fringes, and the paths made equal. The glass plate is then in-
serted in one of the paths, and that path is shortened, or the other path
lengthened, by moving the mirror, till the sharpest sodium maxium is
reached. The path containing the glass plate will be shorter than the
other by about one-half the thickness of the plate. If white light is now
substituted, the fringes should be in the field. It is better before throwing
on the white light, to re-adjust the mirrors for perfect parallelism of the
two beams, which may be known by the almost startling distinctness of the
sodium fringes in this position. It will be easier if the first trials are made
with a plate not more than 5 mm. thick.
With a thick glass plate in one path, the fringes are not all in the field
at one time. It is necessary to vary the adjustable path by a millimeter
or more, depending on the thickness and dispersion of the glass, in order to
run through the range. Toward one end of the range the fringes are al-
ternate circular bands of dark green and bright red. Toward the other end,
the bands are bright green and dark red. In the middle of the range
the bands are almost colorless, the bright circles being yellowish, and the
dark ones a bluish gray. There is no determinate central fringe; the
changes in coloration are so gradual that no difference can be seen in pass-
ing over even as many as a hundred fringes if the glass is fairly thick.
The following table gives the results of a few trials in counting the fringes
with different thicknesses of glass. The figures are not strictly compara-
ble among themselves, as the plates were not all of the same kind of glass.
VoL. lo, 1924 PHYSICS: W. N. BIRCHB Y 453
Thickness of glass plate in mm. 0 3.18 7.44 15.04 30.08
Number of fringes counted 20 255 555 1490 2500
The bands are so uniform as to suggest monochromatic light, though
alternate bands are almost complementary in color. With 7.44 mm. of glass
in one path, the distances passed over in counting 100 fringes at each ex-
treme were measured, and the equivalent wave-lengths calculated. These
were 6360A and 5180A for the bright red and the bright green ends,
respectively. These are merely rough measurements, made to ascertain
the general nature of the fringes. In counting fringes in these experiments,
the adjustable path was slowly lengthened (or shortened) and the number
of fringes vanishing (or forming) at the center counted.
The fact that interference effects are still visible, in spite of the great
dispersive action of a thick glass plate, is most unexpected, and makes an
explanation a matter of particular interest. The theory is given for plane
wave-fronts, and is divided into two parts: I. Wave-front parallel to
the mirrors. II. Wave-front oblique to the mirrors. The present paper
deals with the first part. A paper covering the second part is in prepara-
tion. It should be noticed that the first part does not deal with the whole
field of view in the interferometer, but only with the central spot of the
fringe system.
Physical Explanation of the Ftinges.-To analyze the phenomena, let us
suppose at first a source of monochromatic light. The phase-difference of

4)2Lo x1 qTr qS-0 Or 471-

I I I I I tNW). 11 I1 I I1 - 1 I I.
:D
iso W 47r at
I v r 1/ P 1-
vT d
I
C. I
*1 sa r v v v \.'M
\I I

T
-1-

.,-
A
1-
,1f-A"A.1 Al -A
\%'1Iv,I
A. A. 1- - -1
N,'1
4'I A
NA NA
I A 11 I I
1
.

\ -N A \1
4t *uso LT #g
e.
t
X II /1///X FIGURE
fI.
I d1NNN
1

the two re-uniting beams is proportional to their path-difference, so that, if

i

we have a field over whichthe difference of path increases steadily from left
to right, the points where the phase-difference is an even multiple of -r
454 PHYSICS: W. N. BIRCHB Y PROC. N. A. S.
(greatest light intensity) will be a series of equidistant vertical straight
lines, the distance between them being proportional to the wave-length of
the light employed (fig. 1A). If, now, we place a retarding medium in
part of the shorter of the two paths, it will be necessary, if we wish to restore
the original phase-difference, to lengthen the longer path. That is to say,
any particular phase-line in the diagram will be moved to the right. The
effect will be to move the system of lines as a whole to the right by an
amount depending on the retarding effect of the medium (fig. 1B).
Now, if we suppose a source of light of a definite number of discrete wave-
lengths, we shall have a diagram like the above for each wave-length.
We can graph these one under the other as in figure 1C. Introducing the
retardng medium will move each system to the right by a different amount,
giving an effect shown in figure 1D, where the heavy lines are the positions
of zero phase-difference in each case. Passing to the case of a light-source
containing all wave-lengths, the systems of short lines merge into continu-

1 2tr = 4540 460 4 0 $40o

52oo-

D=IlVStm?. so w2 8
FIGURE 2
9o 19Z 9t 9,6

ous curves, and the phase-difference diagrams appear as in figures 1E and

1F, the former having no retarding medium. The effect of putting in a
greater thickness of retarding medium is to warp the phase-lines more and
more, so that we get the effect shQwn in figure 2, in which the point of zero
path-difference is far to the left of the part drawn. The last mentioned
diagram is drawn to scale, to represent the actual conditions when a glass
plate 3 mm. thick is interposed in the path of one of the beams.
To explain the white-light interference effects observed when the spec-
trum fringes are in this condition, divide the diagram into horizontal strips
of finite but narrow width dA. This is equivalent to considering the source
VoL. 10, 1924 PHYSICS: W N BIRCHB Y 4515
of white light as being composed of a number of sources, each emitting light
of all wave-lengths over a narrow range d\. If we observe the light from
one of these sources, excluding all the others, there will be visible interfer-
ence if, for a given path-difference, the phase-difference does not vary much
over the range. If the phase-difference at one edge of the strip differs by
as much as 7r from that at the other edge, varying continuously and grad-
ually from one edge to the other, the variations in intensity over the strip
will compensate each other, and the field will appear uniform to the eye,
since there are no decided color-differences in the small range of X over the
strip. It will be seen that this last condition holds, for a constant path-
difference and a small range of X, over the whole field in figure 2, except for
those strips in the neighborhood of the bend of the phase-curves. In this
region each strip will show clear interference, since the phase-differences
on its upper and lower edges are nearly the same. Moreover, the neigh-
boring strips will reinforce the interference, their phase-differences being
in step with it. Over the rest of the field the action of neighboring strips
will produce no interference effect, giving uniform illumination. We shall
have, then, interference in one narrow range of the spectrum, superimposed
on uniform illumination from the rest of the spectrum. The color effects ob-
served can be explained if we consider the field to be composed of a set of
interference bands of light of a certain color (small color-range), seen
against a background of uniform illumination of all other colors. The
background will be complementary to the interference bands. The
intensity of the interfering color ranges from twice the intensity of the other
colors to zero, since the rest of the spectrum is covered by very narrow
spectrum fringes, cutting down its intensity one-half. When the inter-
fering light is in the red, for instance, we have alternately a surplus of red
and no red, similarly for other positions. The part of the phase-curves
drawn in figure 2 represents less than one-third of the range of D over
which fringes are visible. If the entire range were drawn it would be no-
ticed that the position of the bend in the phase-curves passes in succession
through all values of X, from about 6000 A. at the left, to about 5000 A. at
the right. A study of the equations given in the next section brings out this
fact, which is also in accord with experiment. The dotted line shows the
change in wave-length of the fringe-producing light as D varies.
It will be noticed that, according to the above analysis, the interference
is caused by light of such a wave-length that its phase-difference is greater
than that of any other wave-length, for the given position of the mirrors.
This is in accordance with the principle enunciated by Cormu,** and en-
larged upon and exemplified by Lord Rayleigh,*** that the condition for
white-light interference is that the phase-difference shall be a maximum or a
minimum with respect to the wave-length. In symbols, d-AdX = 0,
where qb is the phase-difference of the two beams.
456 PHYSICS: W. N. BIRCHBY PROC. N. A. S.
Lord Rayleigh gives several methods of increasing the number of white-
light fringes, by means of optical systems which tend to equalize more or
less the widths of the monochromatic fringes of all wave-lengths. In the
case considered in this paper no such explanation of the increased number of
fringes is possible, since the "width of a fringe" is simply the distance through
which the mirror must be moved to change the phase-difference by 2 7r, and
is equal to half a wave-length in each case. Hence all the monochromatic
components have fringe-widths proportional to their wave-lengths.
M Theory of the Fringes.-In this section
we are going to investigate by what dis-
tance the mirror can be displaced with-
out destroying the interference fringes,
p and we shall find an expression for the
number of fringes visible with plates of
different thickness and dispersive power.
L S length N1
Let be the number of waves of
1sC4
'.Z. X (in air) in Pi, (fig. 3), and N2
L the number in P2, P1 and P2 being the
distances from a point on the half-silvered
surface to the two mirrors. Suppose a
0 plate of glass, or other dispersive sub-
FIGURE 3 stance, of thickness t to be placed in P2.
Then, if , is the index of refraction from air into the plate for wave-length
X, we have the relations,
N1 = P1/X; N2 = (P2- t) /X + pt/lX - (1)
Let D be the difference of the distances to the two mirrors, P1 -P2,
and let 4 be the phase-difference of the beams on re-uniting. Then 0 =
47r (N1 - N2), since the beams pass twice over their respective paths.
Directly from these relations we obtain the fundamental equation
0/4XA DD-(A-1)t (2)
To apply Cornu's principle, differentiate (2) with respect to X, and put
d4/d = 0 in the result. We get
4'/47r + t d,/dCa = 0 (3).
as the condition for producing fringes in the neighborhood of X. Elimi-
nating 4 between this equation and (2), we get the condition in terms of D:
D = (A- 1 - X dpl/dX)t (4)
If X1 and X2 are t4e shortest and the longest wave-lengths of the range under
consideration, by placing the corresponding values in the above equation
and subtracting, we have
D2- D = R = [/ -X2(d/dX)M j-ux + Xj(du/dX)jt (5)
Voi. 10, 1924 V ,PHYSICS: W. jYIRCHB Y 4557
If j, is a continuous function of X this can be written

R = tfX dX (6)
The number of fringes in this range will be (01 - q4)/2wl, 4) and 4% being.
the values of q5 in (3) corresponding to X, and Xs respectively. We shall
have from (3),
(1- 02)/2r = = 2t[(d/)- (du/d))] (7)
whence
n 2tJ'dX (8)

By means of Cauchy's dispersion formula, , = A + B/X2, (2) gives the

following relation between D, 4), and X:
D = +/4X4X + (A - 1)t + Bt/X2
plotting D against X, using 4 as parameter, we get a series of curves of equal
phase-difference. A vertical cross-section of this set of curves at any point
will give the composition of the light for that particular value of D, or in
other words, for the corresponding position of the movable mirror. Figure
2 was drawn from this equation, using the values: A = 1.5; B=
57 X 10-10; t = 3mm.
Putting Cauchy's value for IA in (6) and (8), and integrating, gives
R = 3Bt(l/X2 - 1/X2) (10)
and
nt = 4Bt(1/X -1/Xs) (11)
These results show that the range, and the number of fringes visible, are
proportional to the thickness of the glass plate, and also to the dispersive
power of the glass, as measured by B. By experiment it is found that the
interference is lost in the glare of the background for wave-lengths much
less than 5100A, or greater than 6300A. Using these limits, and taling
B = 57 X 10-10, we get, from (10) and (11), for a plate 3 mm. thick-,
R = .063 mm.; n = 227. This agrees well with actual count, as can be
seen from the table at the beginning of this paper.
The writer wishes to acknowledge the kindness and patience of Dr.
Paul S. Epstein in reading the manuscript critically and making many
valuable suggestions.
* A brief abstract of this paper was presented to the American Physical Society,
September 1923. In the meantime Sethi (Phys. Rev., Jan. 1924, pp. 69-74) observed
the same phenomenon independently, though his theory is differenit from mine.
** J. Phys., Ser. 2, 1, p. 293 (1882).
* Phil. Mag., Ser. 5, 28, pp. 77 and 189 (1889).