JULY 1, 1939 PHYSICAL REVIEW VOLUM E 56

Diffraction Theory of Electromagnetic Waves

J. A. STRATTON AND L. J. CHU
Massachusetts Institute of Technology, Cambridge, Massachusetts
(Received May 15, 1939)

It has been shown by Larmor, Kottler and others that the classical method of calculating
diffraction from the Kirchhoff formula in terms of a scalar light function cannot be applied
directly to the electromagnetic field since it takes into account neither the vector character
of the field nor the effect of charges along the' contour of the opening. The field equations are
integrated directly by means of a vector analog of Green's theorem. The results are applied
to the calculation of diffraction of electromagnetic waves from a rectangular slit in a screen of
infinite conductivity. The results are compared with an exact solution obtained recently by
Morse and Rubenstein.

1. INTRoDUcTIQN discontinuous, and the results are finally com-

ECENT advances in the technique of pared to those obtained from the solution of a
generating ultra-high frequency radio waves boundary value problem by rigorous methods.
have stimulated interest in a number of problems
of electromagnetic theory which heretofore have 2. THE KIRCHHOFF FORMULA FOR SCALAR
had only academic importance. A natural conse- WAvE FUNcTIQNs
quence of this trend towards short waves is an Let s (x, y, s) be any solution of
application of the methods of physical optics to
determine the intensity and distribution of V'q+k'(p =0
radiation from hollow tubes, horns, or small
openings in cavity resonators. Now it is well which is continuous and has continuous first
known that the application of the Kirchho8 derivatives throughout a closed domain U, and
diffraction formula to an opening in an opaque let ps, (Bp/Bn)e denote the values of p and its
screen involves several fundamental errors of normal 'derivative on the bounding surface S.
principle. Nevertheless a remarkably good agree- Then if the direction of the normal is outward
ment between measured and calculated intensi- from S, the integral
ties is obtained in the region directly in front of
u(x', y', s')
the opening provided the wave-length is small
e'""
—vs B (e' d+ (2)
"$
compared to the size of the opening. If, on the
other hand, the wave-length is relatively large,
1 t'Bp'r
—
the radiation is distributed over a wide angle
4s. e EBn)e r Bn(r ) I I

and the Kirchhoff formula proves to be definitely represents a discontinuous function which at all
inaccurate. It must then be extended to account interior points x', y', z' is equal to q and at all
first for the vector character of the wave, and external points is zero. Since the values of q
secondly for the discontinuities introduced about and its derivatives at all interior points are
the contour of the opening. Both of these factors uniquely determined by the value of yz alone on
as sources of error have been recognized by 5 (Dirichlet Problem), or by (Bp/Bn) s alone
previous writers. In the present paper we shall (Neumann Problem), the functions s e and
recall first the restrictions on the Kirchhoff (B&p/Bn)s cannot be chosen independently if e
formula when applied to a scalar wave function. and I
are to be identical at interior points. The
A vector analog of Green's theorem will then be function defined by (2) satisfies (1) and is regular
derived, from which an integral representation within the domain U whatever the choice of yq
of electromagnetic fields can be obtained in a and (B p/Bn) e, but the values assumed by u and
very simple manner. Next the theory is extended Bu/Bn on S will in general diRer from those
to surfaces over which the field vectors are assigned to es and (Bp/Bn) s
9
100 J. A. S rRATTON AND L. J. CHU

A Kirchhoff diffraction problem is formulated electric and magnetic vectors. Each one of these
usually somewhat as follows. A primary wave is scalar functions satisfies (1) and its value at an
incident upon an opaque screen in which there interior point x', y', s' is therefore expressed by
is an opening S~. The scalar potential of the (2) in terms of its values over the boundary S.
wave, or. a rectangular component of a field But these components at an interior point must
vector, is represented by y satisfying (1). It is not only satisfy the wave equation, they must
now assumed that on the dark side of the screen also be solutions of the Maxwell field equations
ys and (By/Bn) s are zero except over the The real problem, therefore, is not the integration
opening Si, where ys and (By/Bn)s have their 'of a scalar wave equation, nor even a vector
undisturbed values. The diffracted wave is then wave equation, but of a simultaneous system of
calculated from (2), the integral extending over first-order vector equations relating the vectors
S~ alone. E and H. There is nothing new in this remark.
Such a procedure cannot possibly give an The integration of the field equations in closed
exact result. For in the first place the assumption form has been discussed by Love, ' Larmor, '
of zero values for the light function and its v. Ignatowsky, ' Tonolo, Macdonald, ' Tedone, '
derivative over the screen S2 implies a discon- and most completely by Kottler, ' but their
tinuity about the contour C& bounding the results appear to have been commonly disre-
opening, while Green's theorem, upon which (2) garded in treatises on physical optics. Recently,
is based, is valid only for functions which are however, the subject has been reviewed by
continuous over a complete bounding surface. Schelkunoff' in connection with equivalence
This difficulty cannot be obviated by the theorems.
common expedient of replacing the contour of
discontinuity by a thin region of rapid but 3. A VEcToR ANALoG oF GREEN s THEoREM
continuous transition. The vanishing of q q and
(By/Bn)s on any finite part of S would then The integration of the field equations can be
entail a zero value everywhere. The integral achieved most directly and rigorously by a
represents in fact a scalar wave function which method which is wholly analogous to the treat-
approximates the true intensity to a degree ment of the scalar wave equation. Let U be a
which must be determined by other means. If closed region of space bounded by a regular
the ratio of wave-length to size of Si is small, surface S, and let P and Q be two functions of
the radiation is thrown largely forward and the position which together with their first and
Kirchhoff function may differ by a negligible second derivatives are continuous throughout U-
amount from the assumed zero value over the and on the surface S. The divergence theorem
screen. If, on the other hand, the wave-length is is applied to the vector PXiP'XQ, giving
large, the wave function calculated on the basis
of undisturbed values over S~ will be found to
have values over S& which are by no means
small. It is sometimes suggested' that this
f V
P (PXPXQ)dv= ((PXPXQ(
S
nda. , (3)

distribution over S2 be applied once more to the where n is a unit normal vector directed outward
—
Kirchhoff formula (2) a, method of successive from S. Upon expanding the integrand of the
approximations. Quite apart from questions of volume integral a vector analog of Green's first
convergence, the difficulties of evaluating the identity is obtained,
resulting surface integrals in most cases make
such a procedure of little practical value. ' A. E. H. Love, Phil. Trans. A197, 1 (1901).
There are other difficulties. An electromagnetic
3
J. Larmor, Lond. Math. Soc. Proc. 1, 1 (1903).
4 W. v. Ignatowsky, Ann. d. Physik 23, 875 (1907); 25,
field at a point on the closed surface S is char- 99 (1908).
5 A. Tonolo, Annali di Matematica 3, 17, 29 (1910).
acterized by a set of scalar functions which 'H. M. Macdonald, Proc. Lond. Math. Soc. 10, 91
represent the rectangular components of the (1911);and Phil. Trans. A212, 295 (1912).
' O. Tedone, Linc. Rendi (5), 1, 286 (1917).
F. Kottler, Ann. d. Physik 71, 457 (1923}.
~
M. Born, OPkik, p. 152. ' S. A. SchelkunoR, Bell Sys. Tech. J. 15, 92 (1936).
 DIFF RACTION THEORY
The vectors E and H satisfy
(VXP VXQ —P VXVXQ)dv
fy VXV XE O'E—
=icopJ VX—
J*, (6)
PX~X nba. 4 VXVXH —k'H=i(oeJ*+VX J, (7)
S
where k'= &a'ep. Solutions of (6) and (7) are to
The vector analog of Green's second identity '

be found which are finite and single-valued at

("Green's theorem") is obtained by reversing all interior points of U and at all points on 5.
the roles of P and Q in (4) and subtracting one In (5) let P=E and Q = Pa, where a is a unit
expression from the other. vector in an arbitrary direction and P = e'""/r
The Green's function Q is essentially the vector
(Q VXVXP —P VXVXQ)de
potential of a unit current element. Distance is
1' measured from the element at x, y, z to the
point of observation at x', y', s'.
PX~X — X~X& nd.
S r = [(~' —~)'+ (y' —y)'+ (s' —s)'j'.
The first identity (4) is the logical basis for The following identities are readily verified.
uniqueness proofs inconnection with vector VX Q = VQXa, VX VX Q =ak'/+V(a Vf),
fields. It will be noted that if one place P = Q =E, V X V XP = k2E+i(op J —V X J*.
(4) proves to be identical with Poynting's
theorem. By application of the divergence theorem and
further transformation, it is easily shown that a.
is a factor common to all the terms in (5), and
4. APPLICATION TO THE FIELD VQUATIONS
since the direction of a is arbitrary, it follows that
The field equations in a rationalized m. k. s.
system of units are VX J
i~pJW P+ PVP de
V-
(I) VXE —i&up, H= —J*., (III) V H= p*/p,
(II) V XH+ia&eE= J, (IV) V E= p/e. [i~p(nXH)i/+(nXE) XV''

It is assumed here that the fields are harmonic +(n E)VP —nX J*gjda. (9)
and that all quantities contain the time in the
form of a factor exp ( — idiot). An application of the identity
The field intensity E is measured in volts per
meter, H in ampere-turns per meter, current
density J in amperes per square meter. In free
space po=4ir X10 ' henry per meter, ~0 (I/36'—
— —
r)
V Sf
VX J"Pdii=J)nX J*gda+ I J*XVPd& (10)
V

reduces this to
X 10 ' farad per meter. The medium is assumed
to be homogeneous and isotropic, and of zero
conductivity. The quantities J~ and p* are
fictitious densities of magnetic current and
i~pJ4 J*XV4+ pV4 —
1
ds-
magnetic charge. In normal fields they are zero,
but the arbitrary (and physically impossible) i~p nXH nXE X~ +
assumption of discontinuities in the tangential S

components of E about the contour of an opening (n E) Vg]da. (11) +

can only be accounted for by some such assump-
tion. Currents and charges are related by the Because of the singularity of the function P
equations of continuity, at r=0, the identity (11) holds only when this
point is excluded from V. The procedure is
(V) V J— ia)p=0, V J+ i(op*=0. — familiar to everyone acquainted with potential
 102 J. A. STRATTON AN D L. J. CHU

theory. The point x', y', s' is taken as the center sources are placed equal to zero. Then
of a small sphere of radius rI. The normal over
the sphere is directed out of V, and consequently
radially towards the center.
E(x', y', z') = —
1
iraq JP+
1
p—
Vif dv,
4m y C

—ik) (e'"'/r) (16)

VP = (1/r
H(x', y', s') = — J X Vgd&.
ro, (12) 1

and on the sphere n=ra. The area of the sphere 4~ v

vanishes with the radius as 4mrP, and since
Since the current distribution is given, the charge
(nXE) Xn+(n E)n=E, (13) density can be determined from the equation
of continuity.
the contribution of the spherical surface to the It is known from the uniqueness theorem of
right-hand side of (11) reduces to 4irE(x', y', z').
electromagnetic theory that a field within a
The value E at any internal point of Vis therefore bounded domain is completely determined by
E(x', y', s') the specification of the tangential components
of E and H on the surface. It follows in (14)
1 and (15) that when nXE and nXH have been
i~pJ4 J*—
XV 4+ pVO-di 6xed, the choice of n E and n H is no longer
4m y
arbitrary. The selection must be consistent with
the conditions on a field satisfying Maxwell's
Lie&p(nXH)P equations. The same limitations on the choice of
4~ s
ye and (By/Be)z were pointed out in ft2. The
+(nXE) XVP+(n E)VP]da. (14) dependence of the normal component of E upon
the tangential component of H is equivalent to
An appropriate interchange of vectors gives that of p upon J.
Let us suppose for the moment that the charge
H(x', y', s')
and current distributions in (14) are confined to
1 1 a thin layer at the sur'face S. As the depth of
icoe J*P+JXVit+ p*VP —
dv the layer diminishes the densities may be
4x y p increased so that in the limit the volume densities
are replaced in the usual way by surface densities.
— Li(06(nXE)P —(nXH) XVP
+4~
1
If the region U contains no charge or current
s within its interior or on its boundary S, the
—(n H) Vg]da. (15) 6eld at an interior point is

It will be shown below under more general

circumstances that (14) and (15) satisfy the E(x', y', s') = — i(op(nXH)P
I
4~ s
6eld equations at all points of U and S.
These expressions are essentially equivalent to +(nXE) XVP+(n E)VQ]da. (17)
those obtained by v. Ignatowsky for a closed
surface. If all sources can be enclosed within a It is now clear that this is exactly the 6eld that
sphere of 6nite radius, the 6eld is regular at would be produced by a distribution of electric
infinity and either side of S may be chosen as its current over S with surface density K, a distri-
"
"interior, or S may be closed at infinity. It bution of magnetic current of density K*, and a
may be remarked that (14) and (15) are con- surface electric charge of density p, where
venient expressions for the calculation of a field K= —nXH, K*=nXE, rl = —en E. (18)
directly from a given distribution of current
without the intervention of vector and scalar The values of E and H in (18) are those just
potentials or of a Hertzian vector. The surface ieside the surface S. The function E(x', y', z')
S recedes to infinity and the fictitious magnetic defined by (17) is discontinuous across S. It is
 DIFFRACTION THEORY 103

well known" that the integral assumption of a line distribution of charges or

1 currents about the contour C. This line distribu-
E3(x', y', s') = gVPda (19) tion of sources contributes to the field and only
4m. ~ when it is taken into account do the resultant
suffers a discontinuity on transition through S expressions for E and H satisfy Maxwell's
given by n DEq —s/o, where EES is the difference equations. The calculation of a diffraction pat-
of the values outside and inside. The third term tern from an integral extended over a portion of
of (17) does not affect the transition of the a surface only, as has been the customary
tangential component but reduces the normal practice, must necessarily lead to erroneous
component to zero. Likewise the discontinuity of results.
A method of determining a contour distribu-
= — K*XVeda
1
E2(x', y', s') tion consistent with the requirements of the
(20)
4~ s problem was proposed by Kottler. It was shown
in f4 that the field at an interior point in (17)
is specified by n)&E2 ——K*, so that the second
is identical with that produced by the surface
term in (17) reduces the tangential component
currents and charges specified in (18). A discon-
of E to zero without affecting the normal
tinuity in the tangential components of E and
component. The first term in (17) is continuous
H in passing on the surface from zone Si to
across S, but has discontinuous derivatives.
zone S2 implies therefore an abrupt change in
ZMp the surface current density. The termination of
V'XE(x', y', s') = — (n X H) X Veda. (21)
4~ a line of current, in turn, can be accounted for
s
on the basis of the equation of continuity by an
The vector E and the tangential component of accumulation of charge on the contour. Let ds
its curl is zero on the positive side of S; it is be an element of length along the contour in
therefore zero at all external points. The same the positive direction as determined by the
analysis applies to H. positive normal n, Fig. 1. Let ni be a unit vector
lying in the surface, normal to both n and the
5. EXTENSION TO DISCONTINUOUS contour element ds, and directed into zone (1).
SURFACE DISTRIBUTIONS The line densities of electric and magnetic
The results of the preceding section hold only charge will be designated by fT and 0.*. Then
if the vectors E and H are continuous and have Eqs. (V), when applied to surface currents,
continuous first derivatives at all points of S. become
They cannot, therefore, be applied directly to n~ (K& —K&) =iooa, n~ (K&* —K~
) i&oo*, =(22)
the problem of diffraction at a slit. To obtain
the required extension of (17) to such cases, and hence by (18),
consider the closed surface S (surfaces closed at i&oo=n,
(nXH, —nXH1)
infinity are included) to be divided into two = (H2 —Hg) (n&Xn), (23)
zones S& and $2 by a closed contour C lying on S,
as in Fig. 1. The vectors E and H and their i~o*=n~ (nXE, —nXE2)
first derivatives are continuous over Si and = —(E2 —E&) (niXn).
satisfy the field equations. The same is true The vector niXn is in the direction of ds. If
over S2. However the components of E and H S2 represents an
opaque screen over which
E2 = H& —0, the field at any point on the shadow
which are tangential to the surface are now
subject to a discontinuous change in passing side is
across C from one zone to the other. The occur-
rence of such discontinuities can be reconciled E(x', y', s') = —
1 1
VPH& ds —
zG06 4'
with the field equations only by the further
1
H. B. Phillips, Vector Analysis (Wiley and Sons, 1933), [icvp(nXH, )g+(nXEi) XVP
p. 206. In greater detail, H. Poincare, Theoric Mathemati- 4~ si
gue de la Lumiere, II, Ch. VII. + (n. Eg) Vf]da, (24)
104 J. A. STRATTON AN D L. J. CH U

)
h
——V'v when
taking. into account the relation V' ' = h
applied to P or its derivatives. Now

(n X H) Veda
S

PVXH. nda — PH ds. (28)

S2 ~
S C

The line integral resulting from this transforma-

tion is zero when S is closed, but otherwise just
cancels the contour integral in (27). Inversely,
only the presence of the contour integral leads
F IG. 1 . Thee contour on which the fields and their derlvha to a zero divergence and transverse waves at
tives are discontinuous. The lower n y' g
zone SI should be nI. great distances from the opening S,. From (II)
and the relation k'/i(dE= i(d—
p follows immedi-
whic h can be shown to be identical with Ko ttler's ately the result
resul V' E(x', y', s') = O.
i
4vrE(X', y', S') = —
z(dE
—
Vt)'dH& dS An identical proof holds for H(x,I y,I ~f
Finally it will be shown that (24) a.nd (26)
.
QE) Xds — (
E) — —P
an)
~da. (25)
satisfy (I) and (II).
s, & an
V'XH(x', y', s') = — —
1
[(nXH)) VV&
For the magnetic field one obtains 4~ sI
4 H(x', y', s') +kEQ(nXH)) —iH)E(n XE&) X Vp]da, (30)
1 ~ since the curl of the gradient is identically zero.
VQE) ds+J~
~
[iddE(nXE&)((d
~

zco p, Furthermore
—(nXH)) XVP —(n ~
H~)VQ]da

f
(nXH~) Vvgda
1 SI
~
VVE, ~
ds+ VH, Xds
Zcvp C
(n vv)~XH))veda
ap aH, ) SI
(26)
s ( an an )
It
remains t o bee shown
s that the fields ex- SI
(n VXH)Vdd f, nsXV (HVQ)ds
SI
pressed by these integrals are in fact divergence-
—( ~
EdVd, s
less and satisfy (I) and (II). Consider first the
divergence of (24) at a pomt x, /
y,I s . c
VVH, ds~

J (nSI
~
)

V' E(x', y', s') h onl . The

the operator Vv~ acts on Vt)t only.
last integral takes account of the fact thaat thee
i 1
—Jf idden(nXH)) field equations are by hypothesis satisfied on Si.
v'pH) ds+ ~
I
~
vp
~

zcv~ 4x c 4~ s Then
+ (n E)) V'f]da V'XH(x', y', z')
k' i k'
PH, ds — ~ —j~ (n ~
E,)ada
1
VPH) ~
ds+
~

zG06
~

[i()n(n X sH)))is
~

zME O'Ir 4m c 4m J
+ 4m — (nXH)) SI
Vgda, (27)
= —iH)EE(x',
+(nXE))XVQ+(n E))V4]«
y', s'). (32)
 D I FF RACTION THEORY 105

The validity of (I) is established in the same

manner.

6. APPLICATION TO A RECTANGULAR SLIT

In Fig. 2 a rectangular slit is shown in an
infinite, perfectly conducting screen coinciding
with the xy plane. Plane waves coming from
the left are incident on this slit at various angles.
The n rmal to the slit surface is drawn to the
left and the intensity of the field is to. be calcu-
lated at any point whose coordinates are R', 0', p'
lying to the right of the screen. Then at suS-
ciently large distances Fi|-. 2, Coordinate system for calculating the diRraction
by a rectangular slit.
r~R' —x cos p'sin 8' —y sin p'sin 8'. (33)
Consider first the case of a normal incidence, tangential component of E must be zero at
8' = ~/2, while the magnitude of the initial
with the electric vector polarized along the x axis
normal component is doubled. In the present
and of unit intensity. The field at great distances
from the slit obtained by evaluating (25) is instance these conditions are expressed by
Eg = Ei '(8') +Eg '(~ e'), —
4n. Eg '= —ik(1+cos 8') cos y'A, (36)
E„.=E„'(8') —E„'(s.-8'),
4~E„'=ik(1+cos 0') sin y'A, (34)
where Eg and E~ are normal and tangential
Eg' —0,
components of the res@/tant field. Applied to
where (34) this gives

sin (~ka cos y' sin e') 4~Eel = —21k cos p A ~

A=4
kcosp sing 4~8„=2ik cos 0' sin q'A.
sin (-', kb sin p' sin 0') ez~~' The energy How or intensity of the diffracted
X (35) wave is proportional to the sum of the squares
k sin y'sin O' R'
of these quantities. The solid curves of Fig. 3
Now if this solution is extended analytically are plots of the intensity in the vertical plane
towards the screen, it is evident that it does not q'=0 for several ratios of slit breadth to wave-
vanish on the plane a=0, as required by the length. Similar curves are drawn in Fig. 6 for
boundary conditions. In fact nothing has been the horizontal plane y'=s/2.
stated as to the location of the surface S2 which In the more general case the direction of
closes the slit surface Sj, and one is free to choose propagation of the incident wave makes an
it in the way which is least liable to violate the arbitrary angle n with the s axis. If the polar-
actual conditions. Thus, in the present case S~ ization is parallel to the x axis, the components
may be closed by a surface S2 which lies just of the incident wave are
behind S~ and over which it is assumed that the ez7z(y zin a+z zoz n) H—( / )~ cos E, '

(3g)
field vectors vanish. This is equivalent to saying
that the field functions (34) can be continued H. = —(e/p)& sin nE,
a.nalytically into the region for which 8')m. /2.
The diffracted field is found from (25) to be
The effect of the screen or bafHe is now taken
into account by assuming it to act as a perfect 4~Eg '= —ik cos y'(1+cos 8' cos n)A,
reHector. The phase relations after reHection are Eg ' —0 (39)
determined by the condition that the resultant 47rEr '=ik sin ip'(cos 8'+cos n)A,
 J. A. STRATTON AND L. J. CHU

/
/ I
/ I

I l
I
]
I
I» /
/

Fio. 3.

Fr@. 7.

/
I
I
I
t
t
I
I

FrG. 4. )s
IT

F&G. 8.
Fedos. 3—8. Polar diagrams of the distribution of field
intensity, showing the diffraction of a plane wave by a slit
in the plane of incidence. The long arrow indicates the
direction of incidence and the short arrow indicates the
polarization of the field at the slit. The numerals on
the curves indicate the width of the slit in wave-lengths
(a/X or b/P ).

and the resultant field to. the right of the plane

FIG. 5. screen obtained by adding the reflected wave is
identical with (37), but with n now entering
into A in the form
sin (-', ka cos y' sin 0')
A=4
k cos q' sin 0'
sin -', kb(sin &p' sin 8' —sin a) e'"s'
X (40)
k(sin p' sin 8' —sin n)

In Figs. 7 and 8 the solid curves represent plots

of intensity in the horizontal ys plane for a=30'
and n=60', The distribution in the vertical
plane is identical with Fig. 3.
FrG. 6. In the alternate case the magnetic vector is
 DIFFRACTION THEORY 107

polarized vertically along the x axis and the remarkably good. It will be noted that the most
incident wave is defined by marked deviation occurs in the immediate
neighborhood of the screen and probably arises
siis(z sin a+z oos a) E— (~/&)'s cos &Il
from the errors which are fundamental to the
(41)
E, = (p/e)~ sin nH, present method: the assumption of unperturbed
distributions over the slit and the manner in
The diffracted field calculated from (25) is then which the reflection problem has been handled.
4irEg '=ioop sin y'(cos 8'+cos a)A, The authors have discarded two points in Figs.
(42) 7 and 8 which fall on the curves published by
47rE, '=ioolz cos y'(1+cos 0 cos n)A, Morse and Rubenstein" and which lead to lobes
for which it is difficult to account. This neglect
and the resultant field after reflection is
may or may not be justified. Professor Morse
4zEg =2tcop sin y cos nA, has kindly put the original data at the disposition
(43) of the authors but insufficient calculations were
47rEy = 2'Loop cos y cos 0 cos 0! c4,
made to settle the matter one way or the other.
where A is defined by (40). The solid curves of It is hard to reconcile the anomalous occurrence
Figs. 4 and 5 represent plots of intensity in the of these lobes with the close correspondence of
horizontal plane. The distribution in the vertical all other points. Whatever the answer to this
plane is identical with Fig. 6. question, the results give strong support to the
It is exceedingly interesting to compare these belief that Eqs. (25) and (26) can be applied to
results with some calculations made recently by the calculation of diffracted radiation with
Morse and Rubenstein, "
who have carried assurance of reasonable accuracy. In the case of
through the two-dimensional problem of diffrac- radiation from hollow tubes and horns, an
tion of an electromagnetic plane wave by an extension to take account of the internal re-
infinite slit. The two methods should lead to flected wave is no doubt possible based on the
approximately the same distribution in an methods employed in acoustics under similar
equatorial plane. The intensity plots obtained circumstances.
from this rigorous solution are shown as dotted » P. M. Morse and Pearl
curves. The correspondence on the whole is
J. Rubenstein, Phys. Rev. 54,
895 (1938).