Waveguides: characteristic modes of hollow rectangular

dielectric waveguides

Katherine D. Laakmann and William H. Steier

The field configurations and propagation constants of the normal modes of a hollow rectangular dielectric
waveguide have been determined. In addition, the coupling coefficients of a Gaussian free-space mode
into the normal modes of a square guide were calculated. The attenuation of each mode is found to be in-
versely proportional to the cube of the guide aperture 2a and proportional to the square of the free-space
wavelength X. For a hollow dielectric square guide with 2a = 1 mm and X = 10.6 um, an attenuation of
0.140 dB/m is predicted for SiO 2 and 0.032 dB/m for BeO. All modes are found to be hybrid modes, al-
though they very closely approximate linearly polarized TEM modes.

1. Introduction The modes for the rectangular guide are analyzed

Hollow dielectric waveguides at optical and ir in the Appendix and the results summarized in Sec.
wavelengths have found wide use in high pressure II. The coupling coefficients for a Gaussian free
laser oscillators and amplifiers. The waveguides are space mode into the normal modes of a square guide
low loss if the cross-section dimensions are many are calculated in Sec. III.
wavelengths and the interior walls have an optical Not surprisingly, the behavior of the square guide
finish. The circular waveguide has been analyzed by in terms of attenuation and mode distribution is
Marcatili and Schmeltzer' and the eigenvectors and much the same as the circular guide, if appropriate
eigenvalues calculated. Recently several groups2 3 allowances are made for the geometry differences.
have built rectangular structures because of the ease The basic differences seem to be that (1) the rectan-
of obtaining polished internal walls and the ability to gular guide supports only EHnm waves and has no
lend themselves to transversely excited discharges. counterpart to the TEam and TMom circular modes,
The rectangular structures are fabricated from four (2) the dominant mode in a square guide is always
polished slabs of dielectrics fitted together to form a the EH 11 mode for any value of dielectric constant.
hollow rectangular structure. Because of the pol- The structure supports low loss modes because the
ishing difficulties of many materials (BeO, A12 0 3), energy propagates essentially within the tube, bounc-
the rectangular waveguide can in some cases be made ing at grazing angles to the walls. Little energy is
superior to the circular structure. transmitted into the dielectric because of the high re-
The guiding structure may consist of a hollow me- flectivity of grazing angles.
tallic duct, a hollow dielectric duct, or perhaps a duct
11. Mode Analysis
made of two dielectric walls and the two metallic
walls. In the optical region of the spectrum, metals Consider a hollow rectangular waveguide of cross
no longer act as good conductors, and a dielectric or section 2a X 2b as shown in Fig. 1. The complex di-
metal may be generalized to a dielectric with a com- electric constant of the surrounding media is Eafor
plex constant. Two dissimilar wall materials are in- two walls, Eb for the other two, and EO,the permittiv-
cluded in the analysis to allow for consideration of ity of free space for the hole. The permeability of all
TE devices (i.e., hybrid device-half metals, half di- media is assumed to be that of free space ,.
electric). An exact analytical solution for the hollow rectan-
gular guide is not possible because of the difficulty in
matching boundary conditions. A computer solu-
tion, similar to Goell's4 for a dense dielectric wave-
guide, is possible, but an approximate solution to the
hollow guide offers better insight to the modes and
The authors are with the University of Southern California, De- gives the mode parameters with sufficient accuracy.
partment of Electrical Engineering, Los Angeles, California 90007. This approximate solution requires the use of the fol-
Received 12 January 1976. lowing assumptions:

1334 APPLIEDOPTICS / Vol. 15, No. 5 / May 1976

 (1) (nX/4b) <<1, Table I. Attenuation
(mX/4a) <<1. Material Loss-dB/m
(2) The dielectric constants must satisfy the con- Sio 2 0.140
dition BeO 0.032
A12 03 (E11c) 0.160
Al2 03 (EIc) 0.086
jibb> >n
-11/2
1 for x polarized modes,
l - 111/2 »m
Za 4a
For a wavelet with polarization perpendicular to
the plane of incidence, the loss L per bounce for
for y polarized modes, small grazing angles can be given by
-ib 111/2>> nX
eb 4b L L -, | On0)1/2 << 1, (2)
where b = Eb/Eo and a = ea/eo.
and similarly for a wavelet with polarization parallel
(3) The boundary conditions will be matched to the plane of incidence
only along the four sides of the hollow region, and the
mismatch of the fields in the corner regions within - 1)1/2 « 1.
L1s (3)
the dielectric will be ignored. In the above, X is the
free space wavelength, and m and n are integers de- From condition (3) the boundary conditions in the
scribing the mnth mode. two transverse directions (x and y) are effectively
Conditions (1) and (2) and the basic properties of decoupled, that is, the coupling mechanism (i.e., the
these modes may be best understood by considering field magnitudes in the corners) is so weak that it has
any field component as a superposition of traveling little effect on the field behavior inside the guide.
plane waves bouncing back and forth between the di- This assumption is consistent to that of low loss
electric walls. Condition (1) requires that the guide modes. Since the fractional power per unit length
dimensions are so many wavelengths that the grazing radiated into the corner regions is very small, the
angle 4n between the plane wavelet associated with fields in the corner are small and hence have a negli-
the nth mode and the dielectric wall is much less gible effect on the internal field behavior or on the
than a radian. The grazing angle An may be written mode properties.
as These are the same conditions that exist for the
/mX\ rectangular dielectric waveguide as analyzed by Mar-
k k4a I)
catili,5 where it was only necessary to match the
fields along the four surfaces of the dielectric wave-
'Pnm = Y - )< 1, (1) guide. Following this approach we will match the
k \4a fields only along the four sides of the hollow region
where k ky are the transverse wavenumbers, and k and neglect the mismatch in the corner regions of the
= 2p/X. In general, modes in which Spn,m is small dielectric. This leads to solutions that vary sinusoid-
enough to be considerably past Brewster's angle will ally internally and decay exponentially into the di-
yield low loss modes, which leads to condition (2) on electric.
the dielectric constant. Utilization of the three listed assumptions lead to
the solution of the characteristic eigenvectors and ei-
genvalues of a hollow rectangular waveguide as given
in the Appendices. The natural modes of a hollow
dielectric waveguide are all hybrid modes with all
three components of the electric and magnetic field
present. If all terms that vary as X/a to a power
1 2b greater than one are ignored, the modes can be classi-
fied into two types: EHxm (x polarized) and EHYm
(y polarized). The appropriate field configurations
E x and propagation constants are listed:
a~ ~ ~ 2 (1) X-Polarized Hybrid Modes EHnm-
Eb

Parity designation sox= 0, m = even integ,

(o = 7r/2 , m = odd integer,
(o = 0, n = odd integer,
soy= 7-/2, n = even integer.

Fig. 1. Geometry of the hollow rectangular dielectric waveguide. The internal field components are

May 1976 / Vol. 15, No. 5 / APPLIEDOPTICS 1335

 -- i=
j (-) cos + cos Y+ Py)exp(-jkgz), IW =-- 2 Re Re [
1 2
H'j (co) / X /~
Hi = 1 (-) Re [ 1
kg'/4b/
A) sin
2a
x+
~x
X sin Y + by) exp(-jkgz), Re(kg) = -[ - __) 4)n ]- (7)
X 2 4a/ 2 4bJ
Exi =sin (Mr x + fp) + sin Pa /
Inspection of the field configuration for both the
2a / L(~~a- 1/ 2ka\ 2a (4) EH'm and EHYmreveals that all transverse fields ap-
proximately vanish at the boundaries. A finite field
X cos (2- x + x)j| Cos Y+ by) at the boundaries is a consequence of the required
2a I \2b
continuity of all internal field components to all ex-
2
-sin [i -1)1/ kb ( ) ] sin (2 Y + PY)J ternal field components. Indeed, the degree of the
finiteness of the fields at the boundaries is directly
Hyi 112E/2i, related to the loss. If all terms in the expression for
the field components of first order or greater in (X/a)
Eyi = 0, are ignored, both the external fields and the longitu-
dinal internal fields vanish everywhere, and, in addi-
Hi = 0.
tion, the internal transverse fields vanish at the
The propagation constant is given by boundaries. The following field components remain:
1 ,MX', F 2
1 For EH'm modes,
Im(kg)
=-- Re[(- 1)/2]
Ei= ()/ 2
() 2
[ ]
sinI- rsin-
in

1/2
s
Hyi ( 2a / j ( 2b m, even; n, even, (8)
R2(kg)= [ 1 2
XL2 2 4a 2 4bi j (5)
.cos (M2rX)] Icos (nrY)] M, odd; n, odd.
(2) Y-polarized hybrid modes EHynn.

Parity designation: (l°X=0, m = odd integer,

SPx= -xI2, m = even integer,
For EHYmmodes,

Ey= () Hi =
[i -
( 2 ) i
sin
. ( 2b )
n, even; m, even,
n, odd; m, odd.
Py = 0, n = even integer, ( 2a l ( 2b).
cos
BP = 0, n = odd integer.
(9)
Comparison of the modes of the circular and the
rectangular guides shows that the circular can sup-
The internal field components are port TEon, TMon, and EHnm modes, while the rec-
tangular guide can support only hybrid EHnm modes.
Ei = ( b\ cos x + ax) co (b Y+ y) exp(-jkgz This is because only the circular symmetry allows
- 4b) cos yx+1)co 2b /)1 closed field lines in the transverse plane without re-
quiring a longitudinal component.
Hzi
=j )()sin (-2aX s4ain + o= The expressions for the attenuation of the modes
as shown in Eqs. (5) and (7) possess two parts corre-
X sin + (PY exp (-jkgz), sponding to the electric field being normal to one
boundary and parallel to the other. The loss for the
Eyi = cos ( x+ x sin
- j 1 IM(! \] parallel E field varies as
l (a - )1/2 a 2a x}
1/(-1)1/2.
Xsin (_X+ ~x) (6)
( 2a )} This is not surprising if one considers parallel polar-
ization where the denser the dielectric, the more effi-
.fsin (2 Y+ y) + sin ji/2 k )J
2b 1 ~~Lib
- cient the reflection.
The loss for the prependicular E field varies as
X cos Y + Yy)l,
- 1)1/2.
-/(Z

Hxi = Exi For perpendicular polarization, as a dielectric be-

comes denser, the reflectivity will first increase and
Exi= 0, then decrease as the corresponding Brewster's angle
becomes closer to the angle of incidence.
Hxi = 0. J It can also be seen from the attenuation coefficient
The propagation constant is given by that a lossy material does not necessarily mean a high

1336 APPLIEDOPTICS / Vol. 15, No. 5 / May 1976

attenuation coefficient. This is because the reflec- m AmJ 2 = Totol Power into guide
tion at grazing angles from a lossy material can be as
high as that from a loss-less material. Indeed, BeO, 1.0
which has a dielectric constant with a relatively large .9
imaginary part and a small real part has one of the z I~~~~~~~~~A,
12
lowest attenuation coefficients for any considered
waveguide materials. Using data from Abrams3 and
Eqs. (5) and (7), we can calculate the attenuation .7
coefficients of several representative guides. For a .6
1-mm X 1-mm guide at 10.6 gtm, the attenuation is
given in Table I. For a guide made of two walls of .5
BeO and two walls of aluminum, a ossible construc-
tion for a TE laser, the lowest loss mode is polarized .4k-
perpendicular to the metal walls and has a loss of ap- .3
proximately 0.016 dB/m.
For a rectangular guide in which all four walls are .2
made of a very dense dielectric, no low-loss mode ex- .1
ists because of the Brewster's angle problem. This IA I31'
result is in direct contrast to a hollow circular guide
1.0 1.2 1.4 1.6 1.8 2.0
made of a dense dielectric. Such a structure sup-
ports low-loss TE modes, which are low loss because (a/wo)
their E fields are tangential to the walls. Fig. 2. Intensity of coupling coefficients of a Gaussian beam into
several modes of a square guide.
111. Coupling Coefficients into the Guide
Because the input and/or output of waveguide de-
vices generally propagate through free space, it is of
interest to know how a free-space Gaussian beam.
couples into the normal modes of a square guide. In
this section, both the amplitude and intensity of the .94~
coupling coefficients of a free-space Gaussian mode
of unity power into the normal modes of a square .92
guide are calculated. The phase front of the Gauss- .90
ian beam is assumed to be flat at the entrance to the l l l l l l l l l I
guide, and the axis is aligned to the guide axis. The 1.0 1.2 1.4 1.6 1.8 2.0
optimum beam waist, in terms of coupling the maxi- (a/w 0 )
mum power into the EH1 1 mode, is 0.69a. This waist
allows 98% of the energy of the Gaussian to couple Fig. 3. Amplitude of coupling coefficient of a Gaussian beam into
into the lowest loss mode. However, the optimum the HE 1 mode as a function of beam waist.
beam waist in terms of maximizing the ratio of the
power into the EH1 1 mode to the total power into the
guide is 0.74a. The derivation of the coupling ampli-
tudes and intensity is given below, and the results are
plotted in Figs. 2-4.
The normal mode of a circularly symmetric Gauss-
ian mode is given by
CO = exp[-(r/wo)1] (10)
.1
(irW02/2)1/2
A13
where w0 = beam waist. The beam amplitude is nor-
malized so that 0

f "*dxdy = 1.
For low-loss guides, the modes essentially form a -. 1
complete set, and we can expand the input Gaussian
in terms of the normal modes. The amplitude of the
nm mode Anm is given by 1.4 1.6
(a/w 0 )
Anm(WO) = CnX(WO)CmY(WO), (11)
where the amplitudes are separated into x and y Fig. 4. Amplitude of coupling coefficient of a Gaussian beam into
components and given by (for n odd) HE1 3 (HE3 1 ) mode of a square guide.

May 1976 / Vol. 15, No. 5 / APPLIEDOPTICS 1337

 CX(WO) 2
= (2)1/ (woa)-'/ i2exp[- (-)] cos () dx, y= k2-kg 2
jweo

(12) - (ki 03i + gky H3i) sin(kix + sox)

cos(kyiy+ py)exp(jkg3),
or alternatively, WEO

Cx(wo) = (2)2 ( ) 1/2 exp [- (nwo) 2]

where kg = wavenumber in z direction;
)XRe [erf - -1 U1.0 )+ C.C.] (13) k ri = internal wavenumber in x direction;
[ (wo i4a ) ] kyi = internal wavenumber in y direction;
(for n even) Eo= internal complex dielectric constant (free
space);
Cnx(wo) = 0. (14) ^3i = internal electric field amplitude in z di-
A similar expression may be obtained for the y com- rection;
ponent. H3 i = internal magnetic field amplitude in z di-
In Figs. 2, 3, and 4, guide coupling coefficients are rection;
plotted as a unction of the ratio of input Gaussian k = (w 2 AEo)Y/1 2 ;
mean waist size to the guide dimension. ox'so = 0 or r/2.
This work was done at the university of Southern
California and at Hughes Aircraft Company. It was Region a Field Components
supported in part by the Joint Services Electronics Eza= cos(kyiy+ ,oy)
exp(-kxax) exp(-jkg3),
3a
Program through the Air Force Office of Scientific
Research under contract F 44620-71-C-0067 and in Hza = H3a sin(kyiy + sy) exp(-kxax) exp(-jkg3),
part by a Hughes Aircraft Company internally fund-
ed program. Ey = _s2 _k2 (+ WU kyi 3a - kaH3a

X sin(kyiy + (,) exp(-kxax) exp(-jkg3), (A2)

2
k - kg2 (wu Y)
X cos(kyiy+ 5oy)exp(-kxax) exp(-jkg3),

Appendix A: Field Components JWEa ak= ka a)

The decoupling of transverse boundary conditions k 2 - kg k~WEax l
as discussed in condition (3) of Sec. II suggests a si- X sin(kyiy + soy)exp(-kxax) exp(-jkg3),
nusoidal behavior of the transverse coordinates in the Hya = a2 2 (k,: 03 + kH3)
interior of the guide. In addition, conservation of k a - kg k x 3 W ak H
energy requires decaying exponential fields in the ex- X cos(Kyiy + py) exp(-kxax) exp(-jkg3),
terior dielectric. With consideration of required par-
ity and the transverse dependence of field compo- where kxa = region a wavenumber in x direction;
nents, the expressions for the longitudinal fields nat- 43a = region a, electric field amplitude in z di-
urally follow. The transverse fields6 may then be rection;
written in terms of the longitudainal fields. The H3 a = region a, magnetic field amplitude in z
symbols SOx and 'pyare equal to zero or p/2 and are direction;
used to allow either even or odd symmetry. ka = (W2A 1/2;
ea = complex dielectric constant.
Inside Field Components

Eyi =
Ez

k -kg

iWIg
k 2 _kg 2
2
= 03i

X
cos(kyiy+ s,)cos(kix +
Hzi = H3i sin(kyiy + opy)

+ ( g Y
Wu
3i + kiH3')
fx.)

sin(kxix + sox)exp(-jkg3),

cos(kxix+ sox)sin(kyiy + so,)exp(-jkg3),

exp(-jkg3),

(Al)
Region b Field Components

Eyb = b
0-
6zb = 63b

Hzb = H 3 b sin(kix

k
g2
+ -9
WUL
+ spx)
cos(kxix exp(-kkyby)

kyb 3b
exp(-jkg3),
+ 4px)exp(-kkyby) exp(-jkg3),

+ kiH3b)

X cos(kxix + soy)exp(-jkg3 - kyby,), (A3)

-( g ki 63i+ kyiH3i)sin(kxix+ () cos(kAy+ ,oy)'exp(-jkg3), Exbb=kbt 2 ( k -k i 6 3b- kybH 3 b)

wu
X sin(kix + spx)exp(-jkg3 - kyby),
H = 2 wfo 2
k -kg HXb = JWfb 6bg g kgW3b)
/_
3b-
02 kgX Wb
- (kyi 3i - H3i) cos(kxix + sos)sin(kyiy + spy)exp(-jkg3),
w6o X cos(kxix+ sox)'exp(-jkg3
- kb)

1338 APPLIEDOPTICS / Vol. 15, No. 5 / May 1976

 a-b;x -y,
2 Wb 2
HYb 0Yk
.- kg WEbkybH3 blI
kxi,0 3 b -kg
b -a;y -x.
X sin(kxix+ so) exp(-jkg3 - kyby).
Equation (C2) is a quadratic equation in tan(kxia +
Appendix B: Matching of Boundary Conditions soX).Substituting in the identities of Eq. (Cl),into
Aty = b: (C2), one can obtain the following two solutions for
Matching of the fields along the propagation direc- the tan(kxia + 'Px):
tion requires that kXi

6 3 b = 63i cos(kyib + y) exp(+kybb)l, tan(kxia + sx) = . I (C4)

H3b = H3 sin(kyib + so) exp(kybb).
" kx" /
Using the above identities in the matching of the x where
and y components of the magnetic field, one obtains
the following matrix equations: -k _,a' = k,,i' + k 2 (-. - 1).

2 2 2
0- _kk' b-g
b
cot
ly"6ib Y
(py)
k,,
w(\k2 a(kyib +ki
k kIW A^b1 Ak,
Ao~yb+
+ kyi !jJ\/43\\1
-gv +i
i/
(B1)

where
= Eb/EO,
eb
2
2 2
Akb = kb - kg ,
2 2
Aki2 = k - kg .

Aka
Atx =a:
With the same procedure and notation as above,

2
the followingmatrix equation is obtained:
k if k i kg [ kx 2 tan(kxia + YX) + k1 i

kxaAa
k 2t (kxia + Ypx)-Ak
kxi2 kg/
W
kyi
2
k . k, /
\~~~A,
ky
ki2)
2
)i H3
(B2)

Appendix C: Solution of Eigenvalue Equation

The wave equation requires
In order to solve Eq. (C4), one of the following as-
ky,2 + kxi2 + kyi2 W2gfo= k2 sumptions must be made:
ky + kx" + ky, 2 = W2 AEa = k"2 (Cl)
eakka < or -| » (C5)
2 2 2 2 2
kg - kyb + kxi - W AEb kb .
The inequalities of (C5) must yield self-consistent re-
In addition, for nontrivial solutions of Eqs. (4) and sults and indeed can be shown to be equivalent to
(5) to exist, the determinant must vanish. condition (2) in Sec. II.
From Eq. (B2) Using the inequality of (C5) in Eq. (C4), one ob-
ky i k i kki ky I\ tains
Aka2Ak'2 Aka2 aki 2
/
-e kXi<< 1 (C6a)
- [ 2 tan(kxia+sx)+ i 2
ea - 1)1/2 k
LAka Akj i
tan(kx'a + 'Px)=
* 2e , cot(kxia + soxl)- -k] (C2) .~'ia- ) »/2k
Aka 2 ~~~Ak
.2 (C6b)
x1ki '>>.
From Eq. (Bi)
The inequality shown in Eq. (C6) is consistent with
ki kxi k ie- k i
_~~ ~~~~
_ Ea R conditions (1) and (2) of Sec. II. Equation (C6) is a
Akb2 Aki2 Akb2 AMi2 transcendental equation, which can be solved easily if
- [kb2tan(kyb
+Y)+ k] the following substitution is made:
kxia + sox= I(p/2) + ba, (C7)
X [ kb b cot(kyb + so,)- 2]- (C3) where
As expected, Eq. (C3) is identical to Eq. (C2) when ba << 1; is an integer.

May 1976 / Vol. 15, No. 5 / APPLIEDOPTICS 1339

Using (C7) to solve Eq. (C6), there are the following Now the propagation constants can be found by
four classes of solution for ki: substituting in Eqs. (C8a) for kx, and Eqs. (C8b) for
From (C6a), kyi into Eqs. (Cl). In addition, the mode configura-
tions may be found by substituting Eq. (Dla) into
sox= 0;kx' Asa [ 1)1/2ka] Eqs. (Al). This substitution reveals that this solu-
tion yields x polarized EHnm,as discussed in Sec. I.
x .; (2) +mr 1+j1 a ]
sox kx J
1)1/ 1 C a
a L E ) kaJ (Ca Type 2 Solutions
From (C6b), Following the same procedure as in the Type 1 so-

+
lution, it can be shown that the conditions,
opx= ; kxi (r/) [1m+ j I ]
cot(kia + sox)- 0,
sox =~2XaL (Z- 1)1/2ka C tan(kyib + y) - 0,
px =r; kxi =r a l+ j (X 1 ] (C8b)

where m is an integer. Obviously, the same solution will give a consistent eigenvector,
holds true for kyi if b is substituted for each a. 3 AIo [k ](X/a)!], (C9b)
Appendix D: Solution of Eigenvectors
H3'\ OC kvi
and yield y polarized EHYm modes as shown in Sec. I.
The determinants of matrix Eqs. (Bi) and (B2)
have been solved independently to yield eigenvalues
k and k i; however, since they both involve the
same eigenvector (63i/H3 9), they must be made to be
consistent with each other. To order (X/a)2 they

0'(X/a).
(_k(Dla)
may indeed be made consistent, as will be shown in References
the following discussion: 1. E. A. J. Marcatili and R. A. Schmeltzer, Bell Syst. Tech. J. 43,
Type 1 Solution 1783 (1964).
2. P. W. Smith, P. J. Maloney, and 0. R. Wood II, Appl. Phys.
Substituting in Eq. (C6a) solutions [i.e., tan(kia + Lett. 23, No. 9, 524 (1973).
Ox) -> 0] into Eq. (5) will yield 3. R. L. Abrams and W. B. Bridges, IEEE J. Quantum Electron.
QE-9, 000 (1973).
H3' 1eo2 k 4. 0. E. Goell, Bell Syst. Tech. J. 48, 2133 (1969).
5. E. A. J. Marcatili, Bell Syst. Tech. J. 48, 2133 (1969).
Inspection of Eq. (Bi) will show that in order to yield 6. E. C. Jordan, Electromagnetic Waves and Radiating Systems
the same eigenvalue, cot(kyib + oyj)- 0. (Prentice-Hall, Englewood Cliffs, N.J., 1950), pp. 260-261.

CORM-NBS WORKSHOP ON
SILICON PHOTODETECTOR MEASUREMENT PROBLEMS

A workshop on silicon detector measurement problems in the 200 to 1200 nm

spectral range will be conducted at the NBSfacilities in Gaithersburg, Maryland.
This will be a one day session on June 3, 1976 the day following the Spring CORM
(Council for Optical Radiation Measurements) meeting. The program being planned
will involve several invited speakers whowill cover various aspects of the sub-
ject. Theywill also discuss other questions or topics dealing with silicon
detectors that are suggested by the audience.
Those who wish to participate in this workshop or-anyone having questions
about specific details should contact:
Franc Grum Edward ZaZewski
Research Laboratory, Bdg. 82 Room A311, BZdg. 220
Eastman Kodak Conpany - or - Radiometric Physics Section
1669 Lake Avenue .NationaZBureauof Standards
Rochester, N. Y. 14650 Washington, D. C. 20234
(716) 458-1000 x77827 (301) 921-2191

1340 APPLIEDOPTICS / Vol. 15, No. 5 / May 1976